Sourcery

BREAKING: Inside Impulse Space w/ Tom Mueller (FULL TOUR)

Molly O’Shea — Thu, 04 Jun 2026 14:49:46 GMT

SpaceX’s #1 Employee Built the Merlin Engine, Now Building Impulse Space

Tom Mueller, Founder, CEO & CTO of Impulse Space, (aka Employee #1 at SpaceX) gives Sourcery a walkthrough of the company’s Redondo Beach factory, from the avionics clean room to a live rocket engine firing in the vacuum chamber.

→ Listen on X, Spotify, YouTube, Apple

As SpaceX’s founding employee, Tom led development of the engines for Falcon & Dragon and started the origins of what became Starship. His proudest project, the Merlin engine, still flies Falcon 9. In this episode he explains why he left to build the next layer of space infrastructure: moving payloads & cargo around once they’re in orbit.

We tour Mira (the company’s precision maneuvering spacecraft), Helios (the vehicle bringing same-day delivery to space), and the Deneb engine, and get into how Impulse designs and builds: extreme vertical integration, 3D-printed engines, in-house composite tanks, and a “build, test, iterate” loop.

Tom also shares where he thinks the industry is heading, on nuclear electric propulsion, data centers in space, the return to the Moon, and what it was like to work with Elon Musk.

Subscribe now

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

(00:00) Tom Mueller, Founder, CEO & CTO of Impulse Space
(00:49) Inside Impulse Space
(02:32) Avionics Bay floor
(02:59) Building rockets at home
(03:50) Mira and Helios
(08:00) Why he left SpaceX
(09:33) The Deneb Engine walkthrough
(11:42) Testing in Mojave
(12:23) Favorite part of the Engine
(13:30) How it's 3D Printed
(14:21) Why 3D Printing changes everything
(16:54) Finding Talent for COPVs
(17:28) No Modern hardware without software
(19:52) The Mill Turn explained
(22:42) Payload Deck Design
(25:28) Entering the Secret Area
(30:48) Thrust, Flow Rate, and 100 Sensors
(32:13) Collision avoidance in Orbit
(32:57) The Electric Propulsion Chamber
(34:28) Nuclear Electric is the future
(38:49) Data Centers in Space
(40:28) SpaceX and Starlink's Growth
(41:10) Working with Elon
(42:07) If not CEO, then what?
(42:32) Moon matters more than Mars

Brought to you by:

Brex—The intelligent finance platform: cards, expenses, travel, bill pay, banking—wrapped into a high-performance stack. Built for scale. Trusted by teams that move fast. visit → brex.com/sourcery

Turing—Turing partners with frontier AI labs to improve model capabilities in coding, reasoning, tool use, & multimodality, as well as with Fortune 500 enterprises to build & deploy end-to-end agentic AI systems in mission-critical workflows Visit: turing.com/sourcery

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Deel—Deel is the global people platform that helps startups hire, manage, pay, and equip anyone, anywhere. Trusted by more than 35,000 fast-growing companies, Deel is the people platform that just works, so teams can scale without the chaos. Visit: deel.com/sourcery

Public-–Investing platform Public just launched Generated Assets, which lets you turn any idea into an investable index with AI. With Generated Assets, you can build, backtest, refine, and invest in any thesis with AI. Gone are the days of one-size-fits-all ETFs. Try it today: public.com/sourcery

Merge—The leading provider of customer-facing integrations and agentic tools for frontier LLMs, Fortune 500 organizations, and B2B SaaS companies. Visit: https://merge.dev

Inside Impulse Space with Tom Mueller: Engines, Spacecraft, & the Case for Moving Things in Orbit

Tom Mueller, founder, CEO, and CTO of Impulse Space, walked Sourcery through the company’s Redondo Beach factory, from the avionics clean room to a live engine firing in a vacuum chamber. Along the way he covered his SpaceX origins, the spacecraft and engines Impulse is building, how the company designs and manufactures hardware, and where he thinks propulsion and compute are headed.

→ Listen on X, Spotify, YouTube, Apple

From Merlin to Impulse

Mueller’s career runs through the heart of modern rocketry. As the first, yes, very first, employee at SpaceX, he led development of the propulsion systems for Falcon and Dragon and started the origins of what became Starship, which he worked on for his last six years at the company. The engine he is proudest of still flies today. “My proudest development was the Merlin engine, which is currently flying on Falcon 9, the most reliable rocket engine ever developed, and also the highest thrust-to-weight of any rocket engine ever developed.”

Mueller was a well established rocket engineer before beginning his tenure at SpaceX, but it was an amateur rocket project that brought him and Elon Musk together. Today, Mueller continues to push the limits of rocketry outside work hours, building rocket engines in his garage “just to get some pre-production data on things” Impulse is working on.

The thesis behind leaving SpaceX was simple. With launch largely solved, Mueller saw the next opportunity one step further out. “Launch mostly being solved or is being solved, it’s like the next big opportunity is to move all that payload, all that cargo around in space.” Impulse exists to do in-space transportation rather than Earth-to-space transportation, a category he believes extends all the way to the Moon, including landers the company is bidding on.

Mira: The Precision Maneuvering Spacecraft

Mira is the company’s precision maneuvering spacecraft, which Mueller describes is built to move quickly once in orbit. It carries eight thrusters, two on each corner, at roughly six pounds of thrust each, for nearly 50 pounds total. For a spacecraft this size, that is a lot: “That’s good enough to move around in space very fast.”

The vehicle packs a full system into a compact chassis. Inside are three propellant tanks, two oxidizer and one fuel, feeding the thrusters. It carries steerable solar panels that let the spacecraft point anywhere while keeping power positive toward the sun, reaction wheels for precise attitude control, and a single avionics box the team calls “the vault.” A gold multilayer insulation blanket wraps the spacecraft and, as Mueller put it, “when that gets in the vacuum of space, it acts like a thermos bottle and really thermally isolates the spacecraft.”

Mira is built for endurance as well as agility, designed to stay on orbit for five years. Mueller framed the product line around a clear division of labor: Mira is the maneuvering vehicle that operates once it arrives, and a separate vehicle gets it there. “Mira is our space maneuvering vehicle that can stay up there for five years, but then we have a vehicle to get you there, and that’s this,” he said, turning to Helios.

Helios: The Vehicle That Gets You There

Helios is the same-day delivery kick stage that does the heavy lifting to distant orbits. At its core is a large propellant tank, holding 12 tons of liquid oxygen and liquid methane, paired with a high-performance engine producing 15,000 pounds of thrust. Mueller summarized its role plainly: it “basically adds a third stage to a two-stage rocket.”

The capability that unlocks is reach. Helios can take a four-ton satellite from low Earth orbit, a couple hundred miles up, all the way to geosynchronous orbit at 22,000 miles, which Mueller described as “escaping most of the gravity well of Earth.” It is a high-energy transfer that conventional approaches struggle to match on time or cost.

The same performance applies well beyond Earth orbit. On a Falcon 9, Helios can send five tons of payload to the Moon, and Mueller said it can improve payload to Mars “by up to a factor of five.” That versatility is why the company positions Helios as the connective layer of its in-space transportation pitch, the vehicle that moves cargo to where it needs to go.

The Deneb Engine

The engine on the bottom of Helios is Deneb, named, like the rest of the family, for a star. Mueller called it “our supergiant engine,” after the blue supergiant it shares a name with. It produces about 15,000 pounds of thrust, is pump-fed, and runs an ox-rich staged combustion cycle, a notably difficult architecture to develop.

Mueller walked through the cycle in detail. Liquid oxygen off the pump runs through the main valve into a preburner, where it burns with a little methane to make hot oxygen at around 4,000 PSI. That hot gas drives the turbine, providing 1,500 horsepower to the turbopumps, then crosses into the main injector. The fuel, meanwhile, flows through cooling channels inside the combustion chamber before entering the injector as hot gas. He expects the result to set records: “This engine will probably be the highest performing hydrocarbon engine ever flown.”

Performance comes from mixing efficiency and expansion. Deneb uses a very fine injector for high mixing efficiency and a large carbon skirt that converts pressure into velocity. That skirt will glow white hot, running at over 3,000 degrees F, likely “the hottest nozzle that’s ever been flown in space.” The thruster Mira uses, Saiph, has about a 290-second ISP; Deneb is targeting roughly 385 seconds. Mueller sees it as the company’s foundation: “this engine will probably become our prime mover,” scalable in the farther future to a multi-engine human lander for the Moon.

How Impulse Builds

The factory is built around extreme vertical integration, with the machine shop, assembly areas, and test all on one floor so the team can “build, assemble, test, & iterate.” That loop is the operating principle, and Mueller traces it directly to lessons from SpaceX. Two capabilities in particular came in-house early.

The first is 3D printing, which Impulse adopted within its first year. Mueller is unequivocal about its value for engines. “For rocket engines, it’s almost like a cheat code. It’s so easy to make a high-performance rocket engine with 3D printing,” because injectors and cooling passages are small and complex. Rather than machine, weld, and braze, “you just basically draw the fluid passages you want and then put a wall around it and just print it.” Deneb’s injector, combustion chamber, and turbopump parts are printed, many in a proprietary burn-resistant alloy the company calls Novaloy, chosen because it has the strength of Inconel but survives the hot-oxygen environment.

The second is composite tanks. Impulse builds its own composite overwrapped pressure vessels, or COPVs, using an aluminum liner to seal propellant and a carbon overwrap to hold pressure. The payoff is mass: “A properly designed COPV will be about half the mass of a state-of-the-art all metallic tank.” The company brought this in-house because tanks are a major portion of spacecraft mass and because buying them externally can take a year or two and cost hundreds of thousands of dollars apiece. Notably, Impulse provides the full propulsion system, including thrusters, for the Vast space station, and is building high-pressure helium tanks for Helios.

Heavy machining rounds out the picture. Copper combustion liners start as 700-pound forgings and are turned down to about 25 pounds, and the entire Mira aluminum chassis is milled in-house on a machine large enough that, in Mueller’s words, “you can machine a motorcycle on this thing.” Software ties it together: in 2026, he noted, “you can’t have modern hardware without software,” with AI used for coding and design and generative tools driving the 3D printing.

Operating in Orbit

Impulse is not theorizing about on-orbit operations; it is doing them. The company has three spacecraft in orbit, all healthy and tracked by GPS. Last year it performed a rendezvous between two of them, closing to within 1,200 meters, which Mueller called “really damn close in space.”

Real operations bring real surprises. Within the first three or four months of flying its first spacecraft, the team had to perform a collision avoidance maneuver, with warnings coming from the Air Force & commercial trackers like LeoLabs. “It kind of caught us by surprise,” Mueller said, “but we were able to do it and get out of the way.”

Steering Mira is its own engineering story. All eight thrusters light, and the spacecraft steers by off-pulsing them, a technique called pulse width modulation. Pulse one thruster at, say, 80 percent duty cycle and it produces about 80 percent thrust, letting the vehicle turn. Impulse is also developing electric propulsion for station keeping. The logic is conservation: use electric propulsion for routine upkeep so the chemical propellant is preserved “for when the customer wants to make a fast move.”

SpaceX Lessons: Starlink, Iteration, & Elon

Some of the most useful parts of the conversation were about what Mueller carried over from SpaceX, starting with Starlink. He described an internal confidence that ran opposite to the public skepticism. “When we first designed Starlink, we all had spreadsheets, and we’re looking at the return on investment and going:

→ ‘We need to make & fly as many as we can.’”

While critics online insisted the satellites would fail, the team saw the opposite. They believed it would “hockey stick the stock, which it exactly did.”

A second lesson is about iteration. Mueller’s rule of thumb, drawn from both of his major engine programs, is that getting hardware truly dialed in takes repetition. “It took three versions of Merlin to get it to be really tight, and it took three versions also of Raptor,” he said. “I would say it takes really three iterations to get to a really, really tight product.” It is a philosophy that maps directly onto the build-test-iterate loop the Impulse factory is designed around.

On working with Elon Musk, Mueller was warm and specific. “Elon’s great to work with. Generally, he just energizes you. He’s got this infectious enthusiasm and gets you just hyped up to go do crazy stuff.” The defining traits, in his telling, were urgency and the ability to draw out people’s best work, along with a knack for recruiting. “He’s also really good at finding and recruiting talent.”

That recruiting instinct points to the cultural throughline. For Mueller, the central ingredient is the team: “the key to success is having a great team.” Impulse, he said, runs the same playbook he learned at SpaceX, a model he summarized in six words. “Vertically integrated, hire the best, and move fast. That’s what we learned at SpaceX.”

The Long View: Nuclear, Orbital Data Centers, & the Moon

Mueller is candid about where he thinks propulsion goes next. He is a strong proponent of nuclear electric propulsion for the outer planets, pointing to the SR1 space reactor mission to Mars as a starting point. His favorite illustration is New Horizons, which flew past Pluto so fast it captured only a few hours of images.

A large nuclear electric stage, he argues, could get there faster & still slow down to enter orbit, then run instruments for decades on reactor power. Closer to home, he expects solar electric to win on cost & performance for cislunar work as space solar panels get lighter.

His bigger bet is that compute itself should move off the planet. “Compute needs to move to space,” he said, calling it his hot take. The logic is clean: “All you need as an input is power, and all you have as an output is data. So you move it to space, you have all the power you would ever need,” beaming results back down a laser link. With power for compute growing more than 15 percent per year, and projections that it could rival all of Earth’s current power use within a few decades, he sees moving it off Earth as a way to relieve a resource crunch rather than wait until it becomes one.

It is worth being precise about what “data centers in space” actually means, because the phrase invites the wrong mental image. As investor Gavin Baker has framed it on the Invest Like the Best podcast, the right picture is not a Pentagon-sized building lifted into orbit but server racks in space, individual racks placed where sunlight is constant and cooling is effectively free, linked to one another & to the ground by laser.

That reframing matters for a company like Impulse, because racks in orbit still need to be placed, maintained, and moved, which is precisely the in-space mobility problem its vehicles are built to solve.

That long view circles back to the Moon, which Mueller considers more important than Mars in the near term. He cited the goal of building and staying at a lunar base, and tied it to resources: the world is projected to face copper shortages, and “we need to start getting copper from the moon, or from the asteroids,” where the supply is effectively limitless. It is the same throughline that runs through the whole conversation, using the resources and the room of space to take pressure off Earth, with Impulse building the vehicles that move things once they are up there.

PS THEY’RE HIRING!!!

→ Listen on X, Spotify, YouTube, Apple

Subscribe now

Check out our full interview w/ President & COO Eric Romo

→ Listen on X, Spotify, YouTube, Apple

The material presented on Molly O’Shea’s website are my opinions only and are provided for informational purposes and should not be construed as investment advice. It is not a recommendation of, or an offer to sell or solicitation of an offer to buy, any particular security, strategy, or investment product. Any analysis or discussion of investments, sectors or the market generally are based on current information, including from public sources, that I consider reliable, but I do not represent that any research or the information provided is accurate or complete, and it should not be relied on as such. My views and opinions expressed in any website content are current at the time of publication and are subject to change. Past performance is not indicative of future results.

Paid Endorsement. Brokerage services by Open to the Public Investing Inc, member FINRA & SIPC. Advisory services by Public Advisors LLC, SEC-registered adviser. Crypto trading provided by Zero Hash LLC, licensed by the NYSDFS. Generated Assets is an interactive analysis tool by Public Advisors. Output is for informational purposes only and is not an investment recommendation or advice. See disclosures at public.com/disclosures/ga. Matched funds must remain in your account for at least 5 years. Match rate and other terms are subject to change at any time.

ICYMI: Out of 50K+ Global Startups, 5 Won The $1M Pitch

Molly O’Shea — Wed, 03 Jun 2026 14:26:59 GMT

Brought to you by:

Brex—The intelligent finance platform: cards, expenses, travel, bill pay, banking—wrapped into a high-performance stack. Built for scale. Visit → brex.com/sourcery

ICYMI: Inside Deel’s Pitch Grand Finale

Five Startups, +50,000 Applications, & a $1M Bet on Borderless Founders

How a global tournament backed by a16z, J.P. Morgan, and Citigroup is rewriting the early-stage funding playbook, and what it signals about the launch of Deel Ventures.

On May 18, the Pavillon Vendôme in Paris hosted what might be the most geographically diverse seed-stage demo day of the year. 80 founders, drawn from a global field of more than 50,000 applicants, took the stage at the Grand Finale of The Pitch by Deel. By the end of the day, 5 had been selected to enter investment discussions with Deel’s venture capital partner network, with up to $1 million per company on the line.

The winners: Zeely AI (Ukraine), Smart Bricks (US & UAE), nybl (UAE), Acceler8 (US), and Alpic (France). They span advertising AI, real estate intelligence, physics-informed industrial AI, workforce planning, and infrastructure for the Model Context Protocol, a portfolio that says as much about where seed-stage innovation is heading as it does about Deel’s investment thesis.

@jpmorgan) started as a competition. What it became was something harder to put into words — a community of builders who refused to let geography be a barrier to funding.\n\n689 ","username":"deel","name":"deel","profile_image_url":"https://pbs.substack.com/profile_images/1957441398292013056/619XdZIG_normal.jpg","date":"2026-05-22T15:57:22.000Z","photos":[{"img_url":"https://substackcdn.com/image/upload/w_1028,c_limit,q_auto:best/l_twitter_play_button_rvaygk,w_88/ngshidgtb0ozau2dabtj","link_url":"https://t.co/BpVnMpg9le"}],"quoted_tweet":{},"reply_count":0,"retweet_count":1,"like_count":14,"impression_count":3625,"expanded_url":null,"video_url":"https://video.twimg.com/amplify_video/2057853026602270720/vid/avc1/1100x720/H9luiZixfgRLlPDK.mp4","belowTheFold":true}" data-component-name="Twitter2ToDOM">

Why This Competition Comes From Deel, Specifically

To understand why Deel is the company running a tournament built around borderless founders, you have to look at how Deel itself was built.

Alex Bouaziz and Shuo Wang met at MIT in 2013 and founded Deel in 2019, joining Y Combinator that same year. The founding insight came out of their own pain: in earlier ventures, both ran into the same wall when trying to hire international talent. The compliance, payroll, and entity infrastructure simply didn’t exist in a usable form. Deel was the answer they wished had been on the shelf.

What followed was one of the fastest scale-up stories in SaaS history:

a16z led the $14 million Series A in 2020
Spark Capital led a $48 million Series B months later
By April 2021, a $156 million Series C made Deel a unicorn
Coatue & DST Global then led a $425 million Series D in October 2021
Emerson Collective came in at a $12 billion valuation in 2022
October 2025, Ribbit Capital led a $300 million round at a $17.3 billion valuation, with a16z, Coatue, General Catalyst, & Green Bay Ventures participating

Now they’re scaling faster than ever, last reported at over $1.4B in ARR. (ehem, this might already be dated.. watch our interview with Alex)

But the operational story is the more interesting one. Deel didn’t internationalize after building a US base. It was global from day one. Forbes famously described a five-person “Navy SEAL” team that traveled country to country incorporating local entities. Today, Deel owns roughly 250 entities globally, runs in-house, in-country payroll teams in 130+ countries, and serves 40,000+ customers across 150+ countries. It is, mechanically, the most distributed HR and payroll company in the world.

That history matters when reading The Pitch. Deel isn’t a US firm dabbling in international sourcing. It is a company whose entire infrastructure is purpose-built for the geographic problem the competition is trying to solve.

The Tournament That Funnels 50,000 to 5

The Pitch was built around a simple premise: capital is concentrated in too few zip codes. Deel’s solution was to run a structured global tournament rather than rely on the warm-intro economy that dominates traditional seed investing.

The structure was unusually compressed. Applications opened in February 2026, and the entire competition wrapped in roughly three months, far shorter than the typical seed fundraising cycle. Seven regional finals were held across Tel Aviv, Dubai, Singapore, New York, London, Berlin, and Paris between March and May, each producing a cohort of regional winners who received a $50,000 SAFE investment and an invitation to Paris.

By the time the field narrowed to the Grand Finale, only 0.1% of applicants remained, a selection rate Deel has publicly compared favorably against top accelerators. Evaluation combined an AI-powered screening system with human expert review, scoring founders on product strength, market opportunity, team capability, traction, and scalability.

The Judges and the Money Behind Them

What gives The Pitch its weight isn’t just the size of the prize pool. It’s the caliber of investors Deel pulled into the room. The Paris judging panel included partners from a16z, Citigroup, J.P. Morgan, and AltaIR Capital, with J.P. Morgan serving as presenting sponsor across the competition. Earlier regional rounds drew in figures including Charles Gorintin (Alan), Cyril Chiche (Lydia), Jean-Claude Le Grand (L’Oréal), Elise Stern (Eurazeo), Jonathan Userovici (Headline), Cécile Mayer Lévi (Tikehau Capital), and Deel’s Executive Chairman & Chief Strategy Officer Philippe Bouaziz.

Deel’s broader partner network for the competition reads like a who’s-who of global tech and finance: a16z, Google, Stripe, J.P. Morgan, Mubadala, Ribbit Ventures, dLocal, Orrick, and Prodware, among others. Several of these names, notably a16z and Ribbit, are also investors in Deel itself, an alignment that gives the competition a built-in coalition of capital with skin in Deel’s broader thesis.

Anish Acharya, the a16z general partner who led Deel’s Series A back in 2020, sat on the Grand Finale panel. His take on the day:

“What struck me about The Pitch was the depth of global talent in the room. Rigorous selection at this scale is rare, & the founders on that stage had so many of the qualities we look for at seed. We’re excited to keep the conversations going.”

That last line matters. The competition’s prize isn’t a check that lands in a bank account on stage. It’s structured introductions to a vetted investor network, with Deel actively shepherding each founder through the conversations.

The Five Finalists & AI

The selected companies share a common thread: they are infrastructure plays, building rails for industries that are about to get reshaped by AI.

Zeely AI (Ukraine). Founded by Dmytro Samoiliuk, Alina Bondarenko, and Yaroslav Samoiliuk. An AI marketing intelligence platform that generates ads, runs campaigns automatically, and scales creative production for SMBs, a segment that has historically been priced out of the kind of performance marketing operations enterprise brands take for granted.

Smart Bricks (US & UAE). Founded by Mohamed Mohamed. An intelligence layer for global real estate investing, building the data infrastructure needed to scale and manage international property portfolios. Cross-border real estate has long been an information-asymmetry problem; Smart Bricks is betting it can be solved with data.

nybl (UAE). Founded by Noor Alnahhas, Mohammed Shono, and four co-founders. Physics-informed AI for the critical industries civilization runs on, think energy, manufacturing, and heavy industry, where standard ML models often fail because they don’t respect the physical laws governing the systems they’re trying to predict.

Acceler8 (United States). Founded by Chinmay Chauhan and Trisha Pathak. AI for workforce intelligence and planning, claiming 5x faster workforce decisions at 5-10x lower cost than the spreadsheets, consultants, and legacy HR systems most companies still rely on. Notably, this is a category Deel itself plays in, making the pick a strong signal of where Deel sees the workforce stack evolving.

Alpic (France). Founded by Pierre-Louis Theron, Frédéric Barthelet, Nikolay Rodionov, Charles Sonigo, and Erica Beavers. A cloud platform built specifically for the Model Context Protocol, the emerging standard for connecting AI apps to data and tools. Alpic lets businesses build and deploy AI apps directly inside Claude or ChatGPT, an infrastructure bet on MCP becoming the way enterprise AI gets distributed.

The Bigger Reveal: Deel Ventures

The Grand Finale also served as the public launch of Deel Ventures, Deel’s dedicated investment arm. The mandate, in Deel’s framing, is straightforward: be an active, strategic partner to emerging startups by unlocking resources that help them grow, and find founders regardless of where they’re based.

Deel Ventures has already made 15+ investments outside of The Pitch competition, signaling that this isn’t a one-off marketing program but a sustained capital deployment effort. The vehicle will continue sourcing and backing seed-stage founders, with The Pitch acting as one, but not the only, funnel.

For Deel, the strategic logic compounds. The company has built one of the world’s largest distributed customer bases (40,000+ businesses across 150+ countries) and runs the local entity infrastructure that early-stage founders typically struggle to set up themselves. That gives Deel Ventures something most operator funds can’t offer: not just capital and distribution, but the actual scaffolding a portfolio company needs to hire and operate globally from day one, the same scaffolding Deel had to build for itself.

Co-founder & CEO Alex Bouaziz framed it this way:

“As a founder, giving back to the next generation of builders is something I care deeply about. The Pitch is how we do that at Deel, by identifying the strongest seed-stage founders in the world, regardless of where they’re from, and connecting them with the people and capital they need to succeed.”

What It Says About Seed-Stage Funding in 2026

A few takeaways worth holding onto.

Geographic diversity is becoming a deliberate sourcing strategy, not an accident. The Grand Finale’s five winners come from Ukraine, the US, the UAE, and France. Compare that to the concentration of AI investment in 2025, when North American AI startups attracted more than $69 billion in venture funding while European and Asian ventures captured roughly $6.4 billion and $3 billion respectively. Tournaments like The Pitch don’t erase that gap, but they create structured access points for founders outside the Bay Area, NYC, and London corridor.

Operator-led capital is filling a gap traditional VC won’t. Deel Ventures joins a growing list of late-stage operator funds that use their distribution, customer base, and brand recognition to source seed deals that institutional VCs miss. The pitch to founders is implicit but clear: you don’t just get capital, you get a partner whose own platform can accelerate your distribution.

AI infrastructure is still the dominant seed thesis. All five of the Grand Finale winners are explicitly building AI, for marketing, workforce planning, industrial systems, or as infrastructure for the MCP ecosystem. One interesting call out is Smart Bricks, which applies AI data intelligence to real estate. The signal: even at the seed stage, the strongest-graded companies in a global field are building the picks-and-shovels of the AI era.

For the five founders walking off the Pavillon Vendôme stage with active investor conversations ahead of them, the prize isn’t the SAFE. It’s the proof point. Out of 50,000, they were the five. That’s the kind of social proof that compounds.

The Pitch by Deel ran from February to May 2026 across seven regional events. Deel Ventures plans to continue investing in seed-stage founders globally.

Want more from Deel? Watch our interview w/ CEO Alex Bouaziz:

→ Listen on X, Spotify, YouTube, Apple

Subscribe now

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Merge—The leading provider of customer-facing integrations and agentic tools for frontier LLMs, Fortune 500 organizations, and B2B SaaS companies. Visit: https://merge.dev

BREAKING: Impulse Space Raises $500M Series D

Molly O’Shea — Tue, 02 Jun 2026 14:40:57 GMT

Why Impulse Raised $500M to ‘Move Things in Space’

Eric Romo, President & COO of Impulse Space, joins Sourcery on the company’s manufacturing floor in Redondo Beach to break down the freshly closed $500M Series D co-led by 137 Ventures and BANNER VC that pushes Impulse past $1 billion raised.

→ Listen on X, Spotify, YouTube, Apple

We get into what the capital actually unlocks (hiring and production, not flashy CapEx), why “space tug” doesn’t cut it for Mira, and how a single successful flight spurred demand from across the industry. Eric explains Impulse’s three products — Mira, Helios, Caravan — and makes the case for staged access to GEO against traditional options.

He also gets candid: reality on orbital debris cleanup, why Impulse is “closer to a fighter jet than a tugboat,” and the hard lesson from his SpaceX days (he was employee #13; CEO Tom Mueller was #1) that made in-housing manufacturing non-negotiable.

Subscribe now

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

(00:00) Eric Romo, President & COO at Impulse Space
(00:58) Fueling the next phase with a $500M Series D
(04:30) Crossing the $1B funding mark
(06:00) The freedom to execute
(07:09) Why startups die from indigestion
(08:30) Building beyond SpaceX
(13:32) How defense became the real opportunity
(16:19) Impulse’s product stack: Mira, Helios & Caravan
(23:43) The economics of urgency
(26:23) "Closer to a fighter jet than a tugboat"
(30:40) The new era of space defense
(37:33) The SpaceX DNA at Impulse
(38:27) The biggest lesson from SpaceX
(44:23) Hiring world class engineers
(45:38) If not COO, then what?
(46:43) The road to first Helios launch

Brought to you by:

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Merge—The leading provider of customer-facing integrations and agentic tools for frontier LLMs, Fortune 500 organizations, and B2B SaaS companies. Visit: https://merge.dev

Impulse Space Series D: An Investor Briefing

Impulse Space has raised $500 million in Series D funding, bringing its total capital raised to over $1 billion. The following briefing shares the founding story, the round, the investors behind it, the product portfolio, the markets and customers it serves, and the company’s stated plans for the capital.

→ Listen on X, Spotify, YouTube, Apple

The Founding Story

Impulse traces back to Tom Mueller’s garage. After nearly two decades at SpaceX leading propulsion, Mueller tried to retire, and, in Romo’s telling, “found himself nights and weekends building rocket engines in his garage.” People he knew well pushed him toward the obvious conclusion: “Hey, look, you’re doing this anyway. Why don’t you have a company around this work?” When you’re Tom Mueller, capitalizing that company is not the hard part. As Romo put it, “When you’re Tom, you know, when you start a company, you can just kinda raise your hand and people then send checks your way.”

That reputation did more than raise money. It pulled in best-in-class technical talent quickly, people Mueller had worked with before, leaving the company well-capitalized and deeply technical almost from day one. Romo himself was the 13th employee at SpaceX; Mueller was, per the payroll system, “employee 001.” The lineage matters because engines are the hard part of any rocket company. “The engines are the thing that fail,” Romo said. “They’re the thing that’s the hardest to build. They’re the thing that’s always the critical path in the schedule.”

The early open question was simply what to build.

Impulse developed the Mira spacecraft on a striking timeline, going “from not existing to being built and ready to fly in about 14 months,” and that first flight set everything that followed in motion.

The Round

The $500 million Series D was co-led by 137 Ventures and BANNER VC. The round was oversubscribed from existing shareholders alone, and Impulse entered the raise from a position of strength rather than need. As Romo described it, “there was no kinda forcing function. You know, it wasn’t sense of urgency or anything around it.”

Romo frames the raise as a higher bar rather than a finish line. “I think some people maybe think that they’re putting numbers up on a scoreboard, and they should be excited about that... That just means the bar just got higher for us, right?” With over a billion dollars now raised, the obligation, in his framing, is to build a business big enough to return it and then some.

The round closed quickly relative to past raises, which Romo attributed to investor demand. He called it “one of the quicker ones I think I’ve ever done just because there was so much interest in the company.”

The Investors

The round was co-led by 137 Ventures and BANNER VC. 137 Ventures, Managing Partner Justin Fishner-Wolfson, was a prior shareholder that stepped up to lead. “Tom helped transform access to space at SpaceX, and now he’s tackling the industry’s next major challenge: in-space mobility,” Fishner-Wolfson said. “Mobility in space is strategic and will define the next phase of the space economy, and Impulse is building the infrastructure to make that possible.”

BANNER VC, Managing Partner Adam Ramada, came in as a new investor on the round. “As activity in orbit increases, in-space mobility becomes foundational,” Ramada said. “Impulse is building the infrastructure that enables the next layer of growth for the space economy.”

Additional participating investors include Founder’s Fund, Lux Capital, and Linse Capital.

Market Growth: $17 Billion → $71 Billion

The thesis behind the round is that mobility after launch is becoming a distinct layer of space infrastructure. Over the past decade launch has grown cheaper and more frequent, but once in orbit, spacecraft have remained constrained, locked into a single orbit or facing slow, expensive transfers. Impulse positions itself as the company building the vehicles and propulsion to move spacecraft after launch.

On the commercial side, the near-term market is geosynchronous communications. Romo described GEO comms as split between low-Earth-orbit players like Starlink and “traditional legacy comms operators” out in geosynchronous orbit, roughly “100 times further away than LEO.” He noted the GEO commsat market “kind of flattened out in the last 10 years,” and part of Impulse’s bet is that better economics for operators can help it grow again.

A second commercial trend is the shift toward smaller, cheaper GEO satellites, a change Romo half-jokingly called the “consumer electronics-ation of GEO spacecraft.” Those smaller spacecraft have historically had poor options for reaching GEO, which is the gap Impulse’s rideshare program targets.

On the defense side, the market is expanding quickly. Space Force’s budget request has grown from $17 billion a few years ago to $71 billion, driven in part by the need to defend assets in higher orbits as more nations gain consistent access to space.

Product: Mira

Mira is Impulse’s precision-maneuvering spacecraft, built and flown in roughly 14 months and now flying multiple missions, executing record-setting orbit changes and autonomous rendezvous and proximity operations (RPO). Its propulsion comes from the company’s Saiph engine.

Mira began with a different thesis. The original idea was a commercial orbital-transfer vehicle, a “space tug” for low Earth orbit, but the economics never worked: a small market, thin margins, and heavy working-capital demands. In Romo’s words, “it just didn’t really pencil.” Everything changed after Mira’s first flight, with the spacecraft demonstrating its ability to precisely maneuver, host and deploy payloads, and re-task on orbit. Conversations with the Space Force sharpened almost overnight, moving from vague possibility to a specific, mutual plan, and the program pivoted toward defense.

Romo is direct that the resulting product is not the tugboat people once imagined. “The product market fit for Mira, for a high thrust, high delta V space vehicle, is it’s closer to a fighter jet than a tugboat,” he said. The mission profile centers on moving rapidly across geosynchronous orbit to get close to and characterize foreign spacecraft, work that is hard from the ground, where an object in GEO can appear as “literally one pixel. It’s a dot.” Impulse is now scaling Mira production to roughly one per month.

Product: Helios

Helios is a high-energy kick stage, effectively a third stage for medium launch vehicles, which Romo sized as “roughly the same size as like the second stage of Falcon 1.” It is powered by the Deneb engine and is scheduled for first flight in 2027.

Helios solves two problems: time and money. A typical geostationary satellite launches into a transfer orbit and takes “somewhere between six and ten months to finally reach their destination,” earning no revenue and burning $4 to $6 million in fuel along the way. The fast alternative, Falcon Heavy, is expensive and flown infrequently in the commercial market. Helios aims to deliver payloads directly to GEO roughly eight hours after launch, at a price below Falcon Heavy.

The value proposition is straightforward for customers, who weigh time saved (often tens of millions of dollars per flight) against hard costs avoided ($5 to $10 million). There is also a design effect: if operators can assume Helios will be available, they can build lighter spacecraft, and Romo noted those savings alone “actually could come close to paying for the ride on Helios.” The product is sold as a service that “looks an awful lot like a launch contract.”

Product: Caravan

Caravan is Impulse’s rideshare program, designed to lower the cost of reaching higher-energy orbits such as GEO. Romo’s analogy: it’s “kind of, you know, SpaceX transporter but for GEO.” Rather than dedicating a flight to one spacecraft, Caravan groups several together, “four, five, seven of them together on the same ride.”

The program targets the small-satellite gap in GEO. Those spacecraft struggle to make slow electric-propulsion transfers work and have had almost no direct-access options; Romo noted there had been “one example in the last two and a half or three years” of small spacecraft hitching a direct ride to GEO.

Demand has run ahead of expectations. “More demand than we expected in that marketplace,” Romo said, with a couple of flights planned around the offering in 2028. As with Helios, availability changes customer design decisions: if operators can count on a rideshare to GEO, “they can go smaller and smaller GEO.”

The Propulsion Family

Underpinning the spacecraft is a family of engines, each tuned to a different mission profile and named for stars. Saiph handles precision maneuvering and orbital repositioning, the work behind Mira. Deneb is the high-energy, long-distance engine that powers Helios. Rigel is built for throttleable applications such as landers and responsive maneuvers.

Propulsion is the company’s technical foundation, the discipline Mueller built his career on and the reason Impulse could move as fast as it has. With several additional spacecraft and propulsion systems in development, the engine family is designed to support missions across both commercial and defense programs.

A defining principle carried over from SpaceX shapes how all of this gets built: in-housing. Romo’s takeaway from the early years was that iteration speed depends on building components internally rather than waiting on outside vendors. “If you wanna move quickly, you absolutely must insource a lot of the development. You must build your own components.” The company pairs that with single-point accountability, what Romo called a “responsible engineering” mindset, telling people, in effect, “You are gonna own this.”

Use of Proceeds

The funding is directed primarily at hiring and manufacturing capacity. Romo was clear this is a people story rather than a capital-equipment one: “people are by far and away the bas- expense of the company and that’s gonna continue.” He estimated 85 to 90% of the budget goes to technical staff.

Impulse has more than doubled its headcount over the past year and is hiring across propulsion, avionics, autonomy, spacecraft systems, manufacturing, and mission operations, with over 200 open roles and more planned. The company now numbers around 500 people.

Its footprint spans facilities in Redondo Beach, CA; Boulder, CO; Washington, D.C.; and a growing test environment in Mojave, CA. As Romo summarized it, “This funding allows us to scale without compromising the quality and speed of execution that define Impulse. Demand for in-space mobility is exceptionally high, and we’re growing our team and production to address it head on.”

→ Listen on X, Spotify, YouTube, Apple

Subscribe now

ICYMI: "We Don't Trust Agents" - Here's How to Control Them

Molly O’Shea — Tue, 02 Jun 2026 12:56:29 GMT

‘Agent Handler’ is now available

Control AI Access, Tokenmaxxed $$$ Bills, & Stop Mass Data Leaks

Merge co-founders Shensi Ding (CEO) + Gil Feig (CTO) join Sourcery for a wide-ranging conversation on building AI infrastructure that quietly powers some of the biggest companies in tech — including OpenAI, JPMorgan, Perplexity, Netflix, Uber, Mistral, and Dropbox.

→ Listen on X, Spotify, YouTube, Apple

Shensi & Gil break down Merge’s three-product suite (Unified, Agent Handler, & Gateway), the make-or-break month that pushed them to rebuild around AI, and why they now believe “English is your programming language.” Gil gets candid on the state of AI security from supply chain attacks, agentic code flooding GitHub, & why the scariest threat is always internal. Shensi shares hard-won hiring philosophy (missionaries vs. mercenaries), her admiration for Benioff and the “beginner’s mind,” and a hot take on companies over-engineering their own models.

Plus: the SaaSpocalypse, the brutal reality of token-maxxing $$$ bills, governing employee AI access, headless Salesforce, & whether today’s AI valuations will ever make any sense at all.

This was a real treat, both of them are more open than other founders, discussing the realities of building a company in the AI era.

Subscribe now

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

(00:00) Shensi Ding & Gil Feig, Co-Founders at Merge
(01:04) Three products. One big bet
(03:20) How Merge made the AI pivot
(04:42) The Classic Innovator’s Dilemma
(05:58) Building culture around AI
(07:10) The leverage nobody’s talking about
(08:52) Codex vs Claude Code
(09:15) The scale nobody knew about
(09:47) SaaS, Finance, and the Biggest AI Labs
(10:46) Why AI companies buy differently
(12:04) What AI sales actually looks like
(13:04) The Fastest sales cycles in the market
(14:35) Why is Cybersecurity broken
(15:59) Merge's solution to agent security
(19:16) Mythos, Wiz, and the GitHub Hack
(22:34) 1,000 Bot signups in one hour
(23:23) Real reason companies pay ransom to hackers
(25:43) The State of AI Infrastructure Costs
(26:41) Internal AI Governance is the next big problem
(29:28) Most Popular Integrations on Merge
(30:54) Big Giants are planning big moves
(31:54) What does Salesforce going headless exactly mean
(33:41) Agents don’t need a UI anymore
(36:59) Can this AI generation actually adapt
(38:25) What Merge looks for in talent
(41:25) The SaaSpocalypse is real
(45:03) Are AI valuations actually insane?
(47:11) How Merge landed OpenAI, Perplexity, Netflix & Uber
(49:02) The Metrics that actually drive the business
(49:58) Biggest misconceptions in tech right now
(51:55) The market is finally catching up to Merge

Brought to you by:

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Merge—The leading provider of customer-facing integrations and agentic tools for frontier LLMs, Fortune 500 organizations, and B2B SaaS companies. Visit: https://merge.dev

Inside Merge: How a Data-Sync Startup Became Infrastructure for Enterprise AI

Merge co-founders Shensi Ding (CEO) and Gil Feig (CTO) recently joined Sourcery for a wide-ranging conversation about how they made a intentional effort to pivot their business for the AI era, the challenges that came with it, the security questions they spend their time on, and why they take a cautious view of AI agents.

Merge has raised $75M from Accel, Addition, & New Enterprise Associates (NEA) to help prevent large-scale data leaks, and much of the discussion traced back to that focus. Here’s a breakdown of the episode, the products behind it, and what it suggests about where enterprise AI is heading.

→ Listen on X, Spotify, YouTube, Apple

The Bet That Reshaped the Company

About a year and a half ago, Merge made a deliberate, three-part push into AI: using it more aggressively internally, repositioning the existing product to sell to AI companies, and launching new AI products. As Shensi put it, “those bets have really paid off, especially in the past year and a half.”

A turning point came from an unlikely source, a paused deal. A single contract that would have taken up roughly 90% of the company’s resources stalled for a month, and Merge used that window to build. “That month actually really transformed the business,” Shensi said, because firsthand experience with AI coding “allowed us to see what was possible with our existing products as well.”

That experience wasn’t delegated. Both founders went back to the keyboard, using Claude Code and Windsurf to ship their second product. Gil framed it as part of the job: “If we are the leaders of this company, we have to know everything there is to know about how AI works, how you build with AI.” It was the innovator’s dilemma in real time. A fast-growing core product was generating revenue, but the founders knew where the space was heading and wanted to shift resources toward where, as Gil said, “the puck is going.”

The Three-Product Suite

Merge sells three products today, each aimed at a different layer of the enterprise AI stack. The original is Merge Unified, which syncs and normalizes data so companies can power experiences like enterprise search and RAG. “You ingest all the data, we normalize it, and it’s there for your customers,” Gil explained. It remains the deterministic backbone many large financial services customers buy first.

The second product, Merge Agent Handler, works as a single MCP server that lets companies offer hundreds of integrations inside agents. It covers internal agents, external customer-facing agents like support bots, and internal use cases. As Gil described it, when a whole team is in Claude Code, Claude, or ChatGPT, “the connectors all just appear in there and work with whatever AI tools you’re using.”

The third is Merge Gateway, infrastructure for switching between hundreds of AI models and routing requests based on policy. It includes a security layer, routing and cost-saving policies, and dashboards. The goal, in Gil’s words, is to help “your whole company to really optimize your AI spend and usage.”

The three map fairly cleanly to different buyers. Shensi noted that large financial services tend to start with the unified API for deterministic use cases, large AI companies lean on connectivity, & traditional SaaS platforms “could be all three.”

Agent Handler & the Employee AI Problem

One of the more useful insights from the episode came from a customer conversation. The CTO of a large financial firm told Gil that employees kept asking why they couldn’t use agents to automate their work the way their friends at other companies could. The answer was regulation: “We can’t let you just connect all of our internal services to all of your external services.” The demand was there. The governance wasn’t.

Agent Handler is Merge’s answer to that gap. It works as a central hub. Employees connect their AI tools to Agent Handler, and Agent Handler connects to every downstream platform in a governed way. As Gil described the pitch to security teams, an employee can use whatever new AI tool they want, but “the only tool you’re allowed to connect it to is Merge Agent Handler.” From there, IT gets visibility into every connection, action, and security violation in one place.

The control extends to granular permissions, which Shensi illustrated with an example. Hire a PR intern who needs to see accounts in Salesforce, and historically you faced an all-or-nothing choice. With Agent Handler, “that IT manager or that CTO can make it so that PR intern can only see accounts and nothing else.” Even if they try to write or delete data, “it’s just not possible.”

Identity sits underneath all of this. Customers want Agent Handler to “tightly couple with our identity providers,” Gil said, so that when an employee leaves, “they immediately get access revoked.” It’s a governance layer built for the reality that employees adopt new tools faster than security teams can track them.

A Cautious View of Agents

Gil’s view on security comes down to a simple distinction. The risk of an agent isn’t its intelligence, it’s its access. He used an analogy to make the point. A capable agent with no tools is like “an evil genius who’s a mass murderer, but they’re locked in a jail cell.” It can’t do much. “The second you connect it to tools, which is what everyone is trying to do right now, that’s where everything goes wrong.”

That view shapes the product. Merge doesn’t rely on asking an agent to behave. “We don’t trust agents,” Gil said. “We can try to set rules, but there need to be hard guardrails and blocks for things like sensitive data being sent across.” In practice, that means scanning outbound payloads in real time and blocking data like Social Security numbers or proprietary information before it leaves a company boundary. It’s the core of the data-leak prevention work behind the company’s funding.

The wider threat landscape supports the caution. Gil pointed to supply chain attacks, where agent-generated code is being pushed to GitHub faster than humans can review it: “Agents are pushing a ton of code. You don’t have enough humans to read all that code, and so things are slipping by.” A single injected vulnerability in a widely-used open source package can spread across many codebases.

The attacks are also getting more capable. Gil mentioned that while on-call that week, Merge saw over 1,000 bot signups in an hour, scanning backend endpoints from IPs around the world. His broader point was that attackers who once faced language and skill barriers now “speak English perfectly,” “write code perfectly,” and “have unlimited manpower all driven by AI.”

How AI Companies Actually Buy

The sales motion has changed. Gil described the old version. A SaaS company would put “seven people who are API experts on the call asking detailed questions around rate limits and how it handles this and that.” The new version is closer to the opposite. AI buyers often show up without that depth, asking for a chart of how things work because, as Gil paraphrased them, “we don’t know the MCP protocol.”

That uncertainty pushes Merge to be more prescriptive. Shensi explained that the team draws on its experience to guide customers toward best practices, since many “don’t have experience with partnerships for these different integrations” and aren’t sure what the best end-user experience looks like. The product has had to adapt to “what people think that they want when no one really even knows what they want right now.”

What buyers lack in specification, they often make up for in speed. Deal cycles for large AI companies move fast because, as Shensi noted, “the competition is just so fierce.” Many skip the proof-of-concept. Gil recalled customers responding to a proposed POC with “we already had our agent build into it. We know it works.”

When they do have specific asks, those tend to cluster around deployment. Shensi said the recurring requests are deployment options, security requirements, and custom connectors, often tied to existing cloud-provider partnerships Merge has to work around.

The Cost Problem & the Case for Routing

For all the talk of “token max,” Gil flagged a less glamorous reality, the bill. Heavy AI usage works well in theory and often in practice, “but then the bill comes to the CFO, and it’s actually really brutal and way worse than they expected.” Even teams that doubled productivity didn’t necessarily double their budget.

This is where Gateway fits in. Merge has “invested really heavily in cost savings” through the product, routing requests across hundreds of models based on policy. Shensi added that margin discipline is becoming a broader industry focus, citing Cursor’s margin improvements as a signal of where things are heading.

The routing logic can be straightforward. For simple requests, Shensi pointed out, “there are a lot of really great open source providers out there too that are very cheap.” Routing a query like “what’s one plus one” or a simple “thanks” to a cheaper model can cut spend, “but it’s really hard to automate that” without dedicated infrastructure.

The takeaway for buyers is that cost control in AI is becoming a real consideration rather than an afterthought. As budgets tighten and CFOs look closely at spend, the ability to route between models becomes part of running a sustainable AI product.

The SaaSpocalypse & Road Ahead

The founders have a clear read on the pressure facing traditional software. Shensi argued that enterprise sales is harder now because “the time to build the same exact product in-house is just significantly lower.” The old negotiating line of “I could just build this” used to carry a real cost. Today that cost is “very, very cheap.”

That connects to a shift both founders returned to throughout the conversation. Programming is becoming more conversational. Gil described vibe coding daily, with AI now “asking clarifying questions” and “building a pretty robust system.” His view: “It gets to the point where English becomes your programming language.” Shensi added that consumer expectations have risen too, so SaaS that can’t deliver automation out of the box will have a harder time competing.

They see the same logic in the move toward headless software, which they discussed through Salesforce’s announcement. The idea is that an agent can do most things via API or CLI without touching a UI. Shensi, who admires Marc Benioff’s “beginner’s mind” approach, noted how difficult it is to sustain a dominant product through multiple tech shifts: “I just would not count Benioff out.”

On talent and staying focused through the noise, their philosophy is consistent. They hire for “missionaries versus mercenaries,” filtering out candidates chasing flashy valuations, and they look for Keith Rabois’ “barrels” who push to get things done over more task-specific “ammunition.”

Shensi’s closing point captured the tone of the whole conversation: “The best way to succeed is to just do things. If you over-intellectualize your company building instead of actually doing anything, you’re too high on Maslow’s hierarchy.”

→ Listen on X, Spotify, YouTube, Apple

Subscribe now

BREAKING: Inside Lead Bank - $56M Investment Now Worth $1.5 BILLION

Thu, 28 May 2026 14:24:43 GMT

Jack Ma, Masa, Jack Dorsey, Jackie Reses - The Story Behind Lead Bank

Jackie Reses has quietly done what almost no one in finance has pulled off — she’s been a Goldman Sachs banker, a private equity partner at Apax, Yahoo’s Chief Development Officer, Executive Chairman of Square Financial Services, a board member at Alibaba, Affirm, and Nubank, and now the founder & CEO of Lead Bank, the $1.5B fintech infrastructure company quietly powering Stripe, Ramp, Walmart, Affirm, Revolut, and over 100 crypto companies.

→ Listen on X, Spotify, YouTube, Apple

In this episode, Jackie sits down with Molly to share the untold stories behind her improbable career — from sitting in Alibaba board meetings with Jack Ma, Joe Tsai, and Masa Son, to brokering one of the largest value-recovery deals in corporate history for Yahoo, to running Square Capital after a 72-hour hackweek project.

She also gets candid about the hype around “de-banking,” what it’s really like inside the White House (her brother is VP JD Vance’s Chief of Staff), how she built Lead Bank to $280M in revenue with zero sales team, and why the future of financial infrastructure is programmable, on-chain, and AI-native.

Whether you’re a founder, operator, investor, or just curious about how the most powerful in American fintech actually work behind the scenes — this conversation is a masterclass.

Subscribe now

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

(00:00) Jackie Reses, CEO of Lead Bank
(01:06) The swiss army knife of Silicon Valley
(02:46) Inside the Alibaba boardroom with Jack Ma and Masa
(05:23) Being one of the only Americans on Alibaba’s board
(10:29) What Jack Ma and Masayoshi Son are really like
(12:32) The biggest lessons Jackie learned from Alibaba
(15:12) From Goldman Sachs to Silicon Valley
(17:44) Why Yahoo hired a PE investor to run HR
(22:53) Yahoo was a hot mess
(25:33) The deal that recovered billions for Yahoo
(26:31) How Jack Dorsey recruited Jackie to Square
(29:53) Three engineers, three days, one crypto platform
(33:51) What Jack Dorsey is really like
(36:13) Being Jack Dorsey’s HR lead during Twitter chaos
(40:03) How to spot real innovation vs hype
(42:21) Why debanking is a myth
(48:10) Why buy a 100 year old bank
(51:23) Growing a bank with no sales team
(54:11) The APIs powering the future of finance
(56:03) How AI is transforming banking
(57:07) The JD Vance connection
(58:39) What it feels like inside the White House
(01:02:53) The biggest misconception about government
(01:04:21) Why Lead Bank’s culture feels different
(01:05:35) The next chapter for Lead Bank

Brought to you by:

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Merge—The leading provider of customer-facing integrations and agentic tools for frontier LLMs, Fortune 500 organizations, and B2B SaaS companies. Visit: https://merge.dev

The Jackie Reses Playbook: Inside the Career of Fintech’s Most Iconic Builder

The Improbable Career Arc: From Atlantic City to the Top of Silicon Valley

Few executives can claim a résumé that spans Goldman Sachs M&A, private equity at Apax, the C-suite at Yahoo, executive leadership at Square, board seats at Alibaba, Affirm, and Nubank, and now the founder’s chair at a $1.5 billion banking infrastructure company, Lead Bank. Jackie Reses can.

→ Listen on X, Spotify, YouTube, Apple

What makes her story remarkable isn’t just the breadth. It’s how unlikely each transition was. She came up in Atlantic City, broke into Goldman’s investment banking division where she won the 1998 Weinberg Award, the firm’s highest IBD honor given to a single person each year.

After Goldman, Jackie spent a decade at Apax Partners, where she founded and led the firm’s TMT group during an era when private equity itself was still emerging as an asset class. She joined early, ran roughly 20% of the firm, and learned to operate in environments that weren’t always welcoming. As she put it on the podcast, the prevailing code was “if you don’t play basketball or go to strip clubs, you’re not cool.” She made it work anyway. By the time she left, she had become one of only five women in the United States leading a private equity firm.

Her secret wasn’t luck. It was the willingness to take seats that looked sideways to outsiders but were strategically central to whoever was paying attention. “My fear with Goldman was that if I stayed there, I’d never leave,” she explained. “I had to go pursue this entrepreneurial dream.” That instinct, to leave comfort for optionality, is the through line of her entire career. She has consistently shown up early to categories that hadn’t yet been named: emerging market private equity, tech driven corporate development, fintech infrastructure, and now stablecoin banking.

The Atlantic City origin story matters here. Jackie didn’t inherit a network or a name. She built one, meeting by meeting, board seat by board seat, deal by deal. And as she now likes to point out, her brother Jacob is currently Chief of Staff to Vice President JD Vance. “I don’t think anyone expected these two kids from Atlantic City to end up where we are,” she said. The improbability is the point.

Inside the Alibaba Boardroom: Jack Ma, Joe Tsai, Masayoshi Son, + Jackie

Few Americans have ever sat on the board of a Chinese technology company. Jackie was one of them, sent in by Yahoo, which owned 40% of Alibaba but had stopped speaking to its leadership. “There was this dynamic where Carol Bartz had been in this contentious fight with Jack Ma that was widely chronicled, and the two companies stopped talking to each other even though Yahoo owned 40% of Alibaba,” she recalled. “I came in and was asked to help clean it up.”

Image: Private photo from Alibaba, with Maggie Wu their CFO (Source: Jackie Reses)

What followed was one of the most surreal assignments in modern corporate history. Jackie flew to Hangzhou multiple times a month, sometimes bringing her young kids, who would sit in conference rooms alongside Alibaba board meetings. She built a working relationship with Joe Tsai, learned to navigate Jack Ma’s vision first leadership style, and partnered with Masa Son on the deals that mattered most. “Joe and Jack and Masa, too, but I spent the most time with Joe, ended up to be wonderful people that I built a relationship with,” she said. The output: tens of billions of dollars in value recovered for Yahoo shareholders through the Alibaba IPO and the subsequent Alipay transaction.

Image: Joe Tsai, Jackie Reses, Jack Ma (Source: Jackie Reses)

Her observations of the three principals are some of the sharpest in the interview. On Jack Ma: “Jack Ma is an incredible visionary. He is very big picture and thinks conceptually about concepts like, ‘I wanna change what middle class looks like in China.’ Well, that’s a pretty bold vision.” On Masa: “Masa would come in and like, ‘Yes, it shall be blessed.’ And I was like, ‘Okay, my deal has now been blessed by Masa.’” On Joe Tsai, she was unequivocal. “Joe Tsai is truly wonderful.”

The deeper takeaway wasn’t about personalities, though. It was about what she learned watching China execute at scale. “When you spend a lot of time in China, you just feel like they have a singular mission, and they will not stop at anything until they achieve that mission,” she reflected. “You really feel like you’re living in the future.” That sense of singular momentum, paired with Jack Ma’s willingness to dream big and just go, became a template Jackie would carry into every chapter that followed.

Yahoo & the HR Job That Wasn’t an HR Job

When Yahoo offered Jackie the Chief Human Resources Officer role, she was confused. “That to me was, like, a total head scratcher,” she said. “I remember thinking, ‘Why in the name of God would anybody hire me to be the head of HR at a tech company?’” On the flight out to California, she worked through the puzzle and arrived at the answer that would define the next chapter of her career: HR at Yahoo wasn’t actually HR. It was capital allocation

“When you’re evaluating how to go build a company, you’re looking at capital allocation. Capital allocation comes in the form of people, and it comes in the form of dollars,” she explained. “That is something that someone in HR could be really good at.” By the time she started, her scope had already expanded well beyond people operations. She picked up M&A, business development, and the search affiliate business, together responsible for roughly a third of Yahoo’s revenue. Within months she was Chief Development Officer, restructuring divisions, divesting assets, and putting the right leaders against the right growth vectors.

Yahoo at that moment was, in her words, “a hot mess.” Activist investors were pressuring the company to sell its Alibaba stake at a discount. Mobile products were neglected. Search had been given away in BD deals that captured no terminal value. “Everywhere you looked, there was, like, challenges across the business lines,” she said. While Marissa Mayer focused on the core business, Jackie focused on Alibaba, Yahoo Japan, and the tax mechanics, which she described as nearly “the largest tax bills ever paid in corporate history.”

The lesson buried in this chapter is one Jackie has applied ever since: the title rarely matches the actual leverage. “People don’t think of HR as an operational job. It really is,” she said. The best operators take seats others underestimate, and then they expand the scope until the seat fits their ambition rather than the other way around.

Square, Jack Dorsey, & the Hack Week That Built a Bank

Jackie’s recruitment to Square (now Block) came as a cold call. “Out of the blue one day, I get a phone call from Jack, who said, ‘I’ve heard about you, and I understand you run HR at Yahoo. Can you explain that to me?’” she recalled. “I was like, ‘I don’t understand this phone call.’” Dorsey was curious about how someone with an operating background ran a people function, and he ultimately convinced her to come to Square, where she would take over a Hack Week project called Square Capital and grow it into a multi-billion dollar lending and banking business.

Image: Block Team (Source: Jackie Reses)

That early team was tiny, “10 colleagues that were pulled together from, the risk team and the finance team, and an operating team,” but the ambition was massive. Jackie ultimately became Executive Chairman of Square Financial Services, helping the company secure an Industrial Loan Company charter in 2020, the first U.S. fintech to do so. Square Financial Services was, in her telling, a category she invented a decade before peers caught on.

The most vivid Square story in the interview is the crypto launch. Jack Dorsey wanted Bitcoin trading on the platform, so he simply made it happen. “He took three engineers, and he sat at the corner of the sixth floor at 1455 Market and didn’t leave the office for three days,” Jackie said. She sat alongside them as cheerleader, pizza orderer, and comic relief.The product shipped that Friday night. The accounting and SEC questions got sorted afterwards.

What Jackie took from Dorsey was the discipline of restraint. “He’ll sit in meetings and not say a word,” she said. “You can either think that is really freaky… or you kinda watch why he’s doing it, what he’s synthesizing, what he’s capturing, and then realize there’s real wisdom in his ability to just zip it.” That, she said, is probably her biggest takeaway from working side by side with him: “Listening really matters and making sure that you’re capturing the essence of what people are saying even when it is an opinion that is divergent from your own.”

Lead Bank: Buying a 100 Year Old Charter to Rebuild Fintech Infrastructure

When Jackie left Square, she knew exactly what she wanted to build. She had spent years watching the world’s biggest fintechs run on the rails of tiny community banks, most of them under $10 billion in assets, with no engineers, in red states. “Imagine being the world’s biggest fintechs and operating on the rails of these tiny banks,” she said. So she went looking for a charter she could turn into something modern. She found it in Kansas City, Missouri, and in August 2022 acquired Lead Bank, a 100 year old community bank, for roughly $56 million.

Image: Lead Bank Team (Source: Jackie Reses)

The premise was simple but ingenious: own the bank, own the holding company, and build the technology layer in house. “In order to do what I wanted to do, you needed to own a bank, and you needed to own a bank holding company,” she explained. “I went and looked across the country for the best banking law to support what I wanted to build, and I came across Lead Bank in Kansas City, Missouri.” From there, she brought in a technical team of product designers and engineers to build a programmable banking infrastructure layer that fintechs, crypto companies, and e-commerce platforms could integrate directly into their products.

In total, Lead has raised $180 million, last valued at $1.5 billion. Counting Andreessen Horowitz, Coatue, Greycroft, ICONIQ, Khosla Ventures, Ribbit Capital, and Zeev Partners among its investors, along with Larry Fink, Larry Summers, and Rob Goldstein personally.

The client list reads like a who’s who of modern fintech: Stripe, Walmart’s OnePay, Ramp, Affirm, Revolut, Flex, Self, and over 100 crypto companies. Last year Lead generated “about $280 million” in revenue and “$31 million of net income,” all without a sales team. “Our clients are really our sales team,” Jackie said. “And between our clients or me, that’s sales.”

The product itself is five core API sets: lending, money movement, card issuing, accounts, and stablecoins. “All of those things sound absolutely boring,” she admitted. “But all of those APIs are built around modern, programmable, interoperable APIs that can be adjusted to build any kind of product that a client needs.” Lead built the global stablecoin card infrastructure for Bridge, now part of Stripe, making it the first U.S. bank to do so.

It’s also one of the only banks in the country capable of running fiat and on chain stablecoin rails concurrently. In a world racing toward agentic commerce and programmable money, that infrastructure is exactly where the puck is going.

The Team Behind Lead Bank

A company built without a sales team has to win on talent density, and Lead Bank has structured itself around exactly that. The founding team came together largely out of Square, where Jackie had worked alongside her current co-founders. “My co-founder, Erica, my other two co-founders, Hamam and Ronak, we all worked together at Square,” she said. That continuity matters. The team had already shipped together, debugged regulatory issues together, and built financial products at scale together before Lead even existed. They knew how to move fast inside the constraints of a regulated environment, which is the rarest skill set in fintech.

Image: Lead Bank Co-Founders Jackie Reses, Ronak Vyas, Homam Maalouf, Erica Khalili (Source: Jackie Reses)

The composition of the company looks almost nothing like a traditional bank. Lead is roughly 60% women, with an executive team split evenly between women and men. “It’s not by purpose,” Jackie clarified. “We just happen to be that way, and we hire the best people for the job.” Her framing on this is sharp and worth repeating. “I think we’re lucky because that means our network includes 100% of the people who are available for the job, and not 50%. And so we get the benefit of having a broad perspective on what amazing looks like.” The advantage isn’t ideological, it’s strategic. Cast a wider net, get better people.

Image: The Moment Lead Bank Founders Heard They Were Approved (Source: Jackie Reses)

That talent strategy extends to the regulatory side of the house, which is where most fintech infrastructure companies break. Jackie has spent more than a decade building relationships across the FDIC, OCC, and Federal Reserve, and her co-founder Hamam helped the FDIC think through machine learning and algorithmic lending as far back as 2016. “From that era where it was the first glimpse into what the precursor to full AI underwriting models look like, I’ve been deeply embedded in a lot of these regulatory environments to build relationships from all levels of the organization,” she said. That institutional memory shows up in how Lead onboards clients, how it structures compliance reviews, and how it ships products that pass scrutiny the first time.

Jackie’s view of the team itself is unmistakably personal. “We have an amazing culture. We have an amazing group of executives, and we’re really lucky that we have a great place to work in,” she said. The reason clients refer Lead to other clients, the reason VCs call Lead the most sought after introduction in fintech, the reason a $280 million revenue business runs without a salesperson, all of it traces back to the quality of the people executing the work. Talent density is the moat.

Architecture & AI: The Five APIs Powering Modern Fintech

The product itself is deceptively simple to describe. Lead is built around five core API sets: lending, money movement, card issuing, accounts, and stablecoins. That framing is intentional. The categories sound old. The implementation is anything but.

The most concrete demonstration of how this architecture works is Lead’s relationship with Bridge, the stablecoin company now owned by Stripe. Lead built the global stablecoin card infrastructure for Bridge, making it the first U.S. bank to do so. It deploys account technology at a scale most legacy banks can’t fathom.

“You could be Bridge, which is now owned by Stripe, and we help build their stablecoin backed cards, and we deploy account technology. And by accounts, I mean literally bank accounts. The only difference is we could spin up a million of them in a very short amount of time versus sitting there with a pen and paper, and hoping for a free toaster at the end of your half an hour sign up for a bank account.”

Lead is also one of the only U.S. banks running fiat rails and on chain stablecoin rails concurrently, with on chain wallets live on Solana.

AI sits inside every layer of the company. Internally, every team uses it. “Our team uses AI in every team in our company. & it doesn’t matter whether you’re in the ops team, you’re in the local lending team, you’re on the HR team. & obviously, our technical teams deploy almost all of their code with AI.” That isn’t a slide in a board deck. It’s the operating reality of how a regulated bank ships product. From compliance reviews to underwriting workflows to engineering velocity, AI is the substrate Lead’s team runs on.

The more strategically interesting use case is what AI enables inside the product. Lead’s APIs are built to be deployed in real time payment flows and agentic commerce environments, where AI agents themselves are executing transactions. “There are core components of our products that deploy AI in the products so that our clients can use them for agentic commerce,” Jackie said. “You could deploy it in a real time payment network or in a checkout modal prompt, and you could use it for agentic commerce, or any type of financial product you’re trying to deploy in context of whatever your product is doing.” That is where the puck is headed. Money movement as code, executed by software agents, settled through programmable bank rails. Lead is one of the very few institutions in the country with the charter, the technology, and the regulatory relationships to make it work.

Crypto, De Banking, & Why Jackie Calls BS on the Narrative

Lead is one of the most active banks in the crypto industry, which gives Jackie a clear view into one of the most heated debates of the past few years: was “de-banking” real? Her answer is unsparing. “I don’t believe there was de-banking. I think it’s a crock of shit.” “There’s 5,000 banks in the United States. We have a lot of red states. Are you’re telling me that in lots of red states, including where my company is headquartered… those banks were not willing to bank, for example, conservative companies?”

Her argument is mechanical rather than ideological. Crypto VC firms funded portfolio companies during 2022 to 2024 by wiring dollars into bank accounts, not by delivering wheelbarrows of cash. “As we go and fund crypto companies of that era, they were being funneled into bank accounts,” she pointed out. What actually happened, she argues, is more nuanced. Some banks pulled out of crypto because they didn’t have enough client density to justify the specialist compliance and finance teams required. “The juice wasn’t worth the squeeze,” she said. “That’s legit. That happened to lots of companies, but that is not systematically de-banking.”

A second category involves clients who weren’t meeting KYC standards. “If you’re not going to comply with US banking regulations, guess what? You’re not going to be able to keep your account at every bank,” she said.

And a third category, banks declining to serve specific executives or industries, falling into reputational risk territory, which she sees as a legitimate business judgment. “Banks can make decisions based on their own reputational risk. Now, whether we all see reputational risk in the same lens, I think that’s up for judgmental debate.”

The framework she uses to evaluate hype across crypto applies more broadly. “Is there a real use case? Does this product actually add durable value?” she asked. “There’s so many things in my industry where you kind of have to scratch your head and say, ‘Yeah, but I don’t understand why that makes something better.’” AI, in her view, sits at one end of the spectrum where mass adoption makes the value obvious. Crypto sits in the middle, with real breakouts emerging from the noise. The discipline is the same: separate momentum from durability.

DC, AI, & What’s Next for Fintech

Jackie’s relationship with Washington runs deeper than most fintech founders. Her brother Jacob is Chief of Staff to Vice President JD Vance, which gives her an unusual window into the federal government and the occasional surreal experience, like flying on Air Force Two from Atlantic City to D.C. after her father’s funeral. “It kinda looks like a business class flight. And I’ve been on many planes that are much nicer,” she joked. But beyond the family tie, she has spent more than a decade building relationships with the FDIC, OCC, and Federal Reserve, including serving on the Economic Advisory Council of the San Francisco Fed.

Her view of regulators is more sympathetic than the industry norm. “I genuinely like our regulators. I think they’re good people,” she said. “The Fed has been largely non-partisan and the nerdiest group of economists you’ve ever seen, really just trying to divine economic wisdom across the United States.” The biggest misconception about working with the government, she argues, is that you can’t have a real conversation. “The more honest and open you are, and square about what your agenda is, the more you can have a conversation that actually provides advice or listens to a perspective about what they’re trying to achieve.”

Her call to action for the AI industry is direct: show up. “There’s so many things that need to be changed to update the world for the AI age, and I don’t think the experts are going to be in the government at all. Therefore, the people who are the experts need to show up in droves and actually provide their insights to the government so that they then can distill the best framework for judgment to build laws.” Lawmakers, in her view, are skilled at writing frameworks, but they can’t write frameworks for technologies they don’t understand. That requires the operators to do the work.

The Future

Inside Lead, AI is already woven through the company. “Every team across our company is using AI,” she said. “Our technical teams deploy almost all of their code with AI.” But the more interesting use case is in the product itself: programmable money movement designed for agentic commerce, where AI agents can execute transactions in real time payment networks or checkout flows. As she put it, “the way that [AI] is built, the way that it’s deployed… you could deploy it in a real time payment network or in a checkout modal prompt, and you could use it for agentic commerce.” The future Jackie is building toward is one where banks aren’t institutions you visit. They’re infrastructure that runs invisibly beneath every modern product.

And she still calls Lead her “baby.” As she said at the close of the interview: “I love Lead so much. It’s like my baby. I don’t think of it as work. I think of it as my favorite thing to do every day when I wake up.” From Atlantic City to Hangzhou to Kansas City, that’s the arc, and the energy, of one of the most consequential operators in modern fintech.

→ Listen on X, Spotify, YouTube, Apple

Subscribe now

ICYMI: Stord Raises $250M at $3B Valuation

Molly O’Shea — Wed, 27 May 2026 18:05:58 GMT

Brought to you by:

Brex, the intelligent finance platform: cards, expenses, travel, bill pay, banking—wrapped into a high-performance stack. Built for scale. visit → brex.com/sourcery

Stord's $250M Series F

A Bet on Physical Intelligence for Commerce

Inside the $250M Series F, the launch of Stord Labs, and the case for physical intelligence as the next durable moat in consumer commerce.

Stord just had a major Series F announcement raising $250 million at a $3 billion post-money valuation (~ 2x its valuation from 12mo ago) in a round led by Strike Capital, with continued participation from Kleiner Perkins, Founders Fund, Franklin Templeton, Baillie Gifford, G Squared, Bond, and Lux.

Alongside the financing, Stord announced Stord Labs, a physical-intelligence lab at its Atlanta headquarters where agentic AI, robotics, and advanced automation are validated against real orders before deploying across nearly 100 facilities.

Subscribe now

The Business at a Glance

Stord operates a fulfillment and consumer experience platform for independent brands. The network powers $15B+ in annual GMV across 1,000+ brand customers and reaches roughly 1 in 4 US households per year. Revenue is approaching $1B, up about 10x over the past 4 years. The first $100M in revenue took 4 years; current run-rate adds more than $100M every 3 months.

The platform spans three layers: a physical network of nearly 100 fulfillment locations; a vertically built software stack (WMS, TMS, OMS, plus a consumer experience suite); and an AI layer that includes StordAI Chat, Search, and Feed. Headcount is 4,000+, with 200+ in software engineering, product, data science, and physical infrastructure and robotics.

Customers named in the round materials include AG1, True Classic, Seed Health, Fanatics, IM8, Hairstory, ALOHA, Monos, American Giant, goodr, Jolie, Native, Fatty15, doingwell, and Climatic Health.

Headline Metrics:

The Thesis

For a decade, the modern stack for online commerce has rested on three abstractions: Stripe for payments, Shopify for storefronts, Meta for discovery. One layer has been left unsolved: what happens after the checkout button is clicked. For 30 years, the post-checkout experience has been the most consequential battleground in commerce, and the one that no infrastructure platform has consolidated.

“The only way to surpass Prime is to build something that goes beyond it, built for everyone else, and that requires infrastructure that, until Stord, did not exist for independent brands.” — Sean Henry, Founder & CEO, Stord

Amazon controls more than 30% of U.S. online commerce, and the reason is not a superior storefront, superior payments, or superior ads.. the reason is Prime: a delivery promise consumers now treat as the baseline expectation against which every other checkout is measured. As Amazon’s relationship with both consumers and the brands it hosts grows increasingly fraught, the opportunity for independent commerce has never been greater.

That expectation gap is the asset Stord is monetizing. For independent brands, the post-checkout experience has been left to a fragmented ecosystem of 3PLs, carriers, and bolt-on software vendors. Stord is collapsing that fragmentation into a single, vertically integrated platform.

Stord’s bet is that consolidating the post-checkout stack (3PLs, carriers, post-purchase software, returns, AI) into a single vertically integrated platform is the only way independent brands can match or exceed the Prime experience.

“Stord exists to level that playing field. We give independent brands the complete commerce stack: the fulfillment network, software, and AI, to deliver a consumer experience that surpasses Prime.” — Sean Henry, Founder & CEO, Stord
“I’ve run over $50M in product sales through Amazon FBA & watched 3 small 3PLs go dark on us with no warning. When we asked Stord to dekit a SKU so we could run a product test, something completely untestable inside Amazon’s system, they did it. The data drove some of the most important product decisions doingwell has made to date. Amazon makes you feel like a SKU. Stord makes us feel like a brand.” — Imran Jawaid, doingwell

Significant Growth

Revenue has compounded at roughly 10x over 4 years. The inflection point in Stord’s growth maps to early 2023, about 6 months after AI became a production-grade tool. Stord’s framing of that timing is that vertical integration across the physical network, software, and data layers positioned the company to apply AI to live operations faster than competitors that owned only one layer.

The software business is the fastest-growing piece of the company. It tripled in 2025 and is growing faster than the overall business. Software bookings more than doubled in Q1 2026, with 7 live products in market.

StordAI was launched in March 2026. As of the round, it has logged 100,000+ unique interactions across 350+ brands. The roadmap includes agentic experiences and an MCP server.

M&A has been a meaningful contributor: 8 acquisitions completed to date, each described by the company as exceeding its targets. Engineering headcount has nearly doubled while product delivery velocity has nearly tripled.

Brand Expansion

The 1,000+ brands on the platform skew toward category-leading independents that compete on the quality of the direct consumer relationship. The mix spans subscription health (AG1, Seed Health, Fatty15, ALOHA, doingwell, Climatic Health), apparel (True Classic, American Giant), eyewear (goodr), CPG (Native), premium home goods (Jolie, Monos), and licensed merchandise (Fanatics).

A consistent theme across customer commentary is that the platform changes how brands operate, not only where they ship from.

“The brands winning in direct commerce today aren’t just fulfilling orders faster. They’re operating with better intelligence. Through StordAI we are giving valuable time back to our team so they can focus on our customers. Stord is the only fulfillment partner we’ve worked with where the technology meets the needs of our global operation. That matters when you’re scaling across markets simultaneously.” — Danny Yeung, CEO and Co-Founder, Prenetics IM8
“Stord’s ability to manage complex kitting, multi-retailer compliance, and high-velocity D2C shipping from a single network gave us the operational backbone to grow without compromise. Meanwhile, we’ve become a top user of their AI features, which has accelerated our business insights and enabled our organization to move faster than ever.” — Josh Bultz, COO, True Classic

“As Jolie has scaled, Stord has been an integral partner in helping us deliver and maintain a premium customer experience while growing our volume so quickly. Their platform and team have allowed us to scale fulfillment across both direct and wholesale channels, improve reliability, and create a much stronger post-purchase experience for our customers. StordAI has added another layer: our team can get answers about inventory and order status instantly, which means we spend less time managing operations and more time building the brand.” — Arjan Singh, Co-Founder, Jolie
“What sets Stord apart is the combination of network scale and technology that’s honestly years ahead of the standard 3PL offering. During a period where supply chain disruptions have become the norm, that infrastructure has given Monos a level of operational certainty we can actually build around. StordAI has made that even more tangible: our team has real-time answers to questions that used to require back-and-forth with ops.” — Victor Tam, CEO and Co-Founder, Monos
“StordAI Chat is a CEO’s dream. I can quickly answer any question like ‘how much inventory do I have on my inbound?’ or ‘how many weeks on hand do we have in this queue?’ StordAI Chat is answering questions people don’t even know they have. It’s really powerful.” — Nick Stachel, Head of Operations, Climatic Health

Stord Labs & the Physical-Intelligence Bet

Stord Labs is the most strategic announcement attached to this round. The premise: the next generation of physical intelligence cannot be built credibly in simulation or vendor demos. It requires real operational complexity at scale, trained against live order data, validated against the same constraints that govern the production network.

Innovations are built and validated against real orders on the same operating system powering production, and proven changes deploy across nearly 100 facilities immediately. The strategic insight is that the moat is not the model; it is the data, the substrate, and the deployment surface.

The numbers underwriting that bet are non-trivial: training models on live fulfillment data across ~100 facilities, $15B in annual GMV, and 8 billion data points per year. The inflection in Stord’s revenue growth (~10x over four years, with the clearest break point coming roughly six months after ChatGPT’s launch) is causally tied to this vertical integration of physical infrastructure, software, and data.

“The largest companies in the world today (SpaceX, NVIDIA, Tesla) operate at the intersection of software intelligence and physical execution. Their advantages compound in a way that pure software cannot replicate.” — Sean Henry, Founder & CEO, Stord

Differentiation & Flywheel

Stord’s differentiation rests on three assets that are difficult to assemble in isolation and harder to assemble together:

A physical network of nearly 100 fulfillment locations (20 Stord-operated, 80 partner sites) running on a single operating system. Hard to replicate with software alone.
A vertically built software stack (WMS, TMS, OMS, consumer experience suite) built by engineers operating inside live warehouses rather than licensed from third parties.
A proprietary data corpus generated by the network itself: $15B in GMV, 8B operational events annually, packages reaching roughly 1 in 4 US households. This dataset cannot be acquired; it has to be generated.

CEO Sean Henry’s framing of the flywheel

“Every new brand that joins the Stord network improves outcomes for every brand already on it, and that compounding effect is showing up directly in how we compete and win.
More volume means better economies of scale for everyone. More facilities means more precise inventory placement and faster delivery times. More data means smarter routing, better demand forecasting, and more accurate delivery promises.
More accurate promises mean more consumer trust, which means more repeat purchases, which feeds more volume.”
“This is not a simple software network effect. It is a physical network effect layered with a software and data advantage, and the combination is what makes it durable. You cannot replicate a physical fulfillment network with code. You cannot replicate years of operational excellence and culture built from the ground up from behind a screen. You cannot substitute the billions of data points across our network continuously training our models. And you cannot hack your way to the tens of millions of U.S. consumers who have seen Stord on their doorstep.”
“The flywheel took a decade to reach the velocity it has today, & it is still accelerating. As the flywheel spins, our advantage accelerates, & is proven by our industry-shattering win rate.”
“The market has historically underestimated the complexity and defensibility of physical infrastructure at scale, and that underestimation is a meaningful part of what has allowed Stord to reach a position that is difficult to replicate. We prefer to be underestimated.”

Market Landscape & Competitive Context

The competitive set sits in three camps:

Traditional 3PLs and asset-heavy fulfillment networks (ShipBob, ShipMonk, Cart.com, legacy enterprise providers). Operational scale but limited integrated software and AI.
Pure-software providers offering OMS, WMS, or post-purchase tooling without a physical network to enforce it operationally.
Amazon FBA, which dominates marketplace fulfillment but by design does not serve independent brand-owned channels.

The lane Stord is consolidating, the integrated post-checkout stack for independent brands, sits between these three groups and has not had a scaled, vertically integrated incumbent.

Two macro factors are relevant to timing. First, warehouse robotics payback periods have compressed to 12 to 18 months in 2026, and operators with vertically integrated data are best positioned to capture that economic shift. Second, agentic purchasing (AI agents originating and completing commerce transactions on behalf of consumers) is likely to favor platforms where software and physical operations are deeply integrated, because agents will optimize on the variables fulfillment controls: speed, reliability, returns, and price.

“Commerce infrastructure gets built once. From our earliest conversations with Sean Henry, it was clear Stord was assembling something rare: software, physical infrastructure, and AI combined in a way that turns fulfillment into a competitive advantage rather than a cost center. We believe the rise of agentic purchasing will increasingly favor platforms where software and physical operations are deeply integrated. Stord is building that infrastructure. That is why Strike is proud to deepen our partnership in this round.” — John Lagomarsino, Strike Capital
“The next defining infrastructure layer has to connect the physical network, the software, and the intelligence layer behind every consumer experience. That’s what makes Stord so compelling. The company combines scaled fulfillment infrastructure, purpose-built software, and AI trained by the patterns of millions of deliveries. We believed in that vision when Kleiner Perkins first backed Stord in 2019, and our conviction has only grown.” — Ilya Fushman, Partner, Kleiner Perkins

The New Era of Supply Chain

Stord has now crossed the threshold from category challenger to category-defining infrastructure. The more interesting reading is that the company is one of a small number of late-stage businesses where the moat genuinely sits at the intersection of physical execution and software intelligence (something I love to cover), and where the data flywheel underwriting that moat is already in motion at the scale required to train the next generation of physical AI.

The question is not whether Stord becomes the consolidated infrastructure layer for independent commerce alongside Stripe, Shopify, and Meta. The question is how large that layer gets, and on what timeline.

“We are The Consumer Experience Company. The other two-thirds of online commerce has its infrastructure now. And we’re just getting into gear.” — Sean Henry, Founder & CEO, Stord

Subscribe now

Brought to you by:

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Merge—The leading provider of customer-facing integrations and agentic tools for frontier LLMs, Fortune 500 organizations, and B2B SaaS companies. Visit: https://merge.dev

BREAKING: Commure Hits $7B Backed by GC, Sequoia, Morgan Stanley

Molly O’Shea — Tue, 19 May 2026 15:31:11 GMT

Saving Doctors 75 Million Hours

Tanay Tandon, Co-Founder & CEO of Commure, joins Sourcery to break down the company’s new $70M round at a $7B post-money valuation, led by General Catalyst with participation from Sequoia Capital, Morgan Stanley, and Kirkland & Ellis.

→ Listen on X, Spotify, YouTube, Apple

With $750M raised to date, Commure now runs inside 500+ healthcare organizations, 3,000+ sites of care, and processes tens of billions of dollars in annual claims — with 85%+ of revenue cycle work completed without a human in the loop.

Tanay started the company at 18 out of his Stanford dorm room. Today, Commure has 1,200 employees across seven offices, supports 200M+ patient encounters a year, and has doubled ARR three years in a row.

We get into the round, the use of General Catalyst’s Customer Value Fund (CVF) for non-dilutive growth capital, why he believes point solutions will be wiped out, the Augmedix acquisition, the Summa Health partnership in Akron, the JPM flip-flops story, and his IPO plans.

Topics covered:

The $70M / $7B round and why they didn’t need the capital
CVF and non-dilutive financing for go-to-market
Building an AI-native OS for healthcare • HCA, Tenet, Epic, and Meditech partnerships
Why platforms beat point solutions
Lessons from acquiring Augmedix
What Alfred Lin, Hemant Taneja, and Teresa Carlson taught him
The path to IPO

Subscribe now

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

(00:00) Tanay Tandon, Co-Founder and CEO at Commure
(01:03) The $70M R&D Sprint
(03:45) Rockefeller's funding playbook beats dilutive VC
(05:42) Tokens killing SaaS margins
(07:14) Going public is the plan
(07:40) Automating healthcare’s back office
(11:02) Scaling Commure’s engine
(11:49) Partnering with Epic
(13:06) Why healthcare point solutions will be dead in 3 years?
(16:42) Learnings from the AWS-CIA deal
(19:13) Going to zero rollouts overnight
(22:23) Nuking malpractice premiums with AI
(23:52) Winning HCA’s trust
(26:27) Saving burned out doctors
(28:24) Buying a bankrupt hospital
(32:30) Ending healthcare interop
(35:41) The viral hiring email
(38:23) Acquiring for distribution
(45:13) Free blood samples?
(47:19) The Stanford hacker who became an enterprise sales rep
(48:21) The cargo shorts fiasco
(53:57) American healthcare isn't broken
(55:28) The agentic healthcare takeover

Brought to you by:

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Merge—The leading provider of customer-facing integrations and agentic tools for frontier LLMs, Fortune 500 organizations, and B2B SaaS companies. Visit: https://merge.dev

Commure’s $7B Bet on AI Healthcare

Commure CEO and Co-Founder Tanay Tandon joined Sourcery to discuss the round, the company’s trajectory, and his vision for rebuilding American healthcare on top of LLMs and agents. Below, a seven part breakdown of the conversation.

→ Listen on X, Spotify, YouTube, Apple

The Funding: A $7B Valuation Without the Cash Pressure

Commure’s latest round is notable less for the dollar amount than for the posture behind it. Tandon was direct about why the company raised:

“This round.. we really didn’t need the capital. We raised it for pricing purposes, and we saw a couple interesting opportunities to accelerate R&D, particularly around AIR, which is our LLM native EMR platform… and our voice agent platform.”

The company has now raised roughly $750 million in total across its life, and the new round was structured to set a new mark without meaningfully diluting shareholders.

A central part of that strategy is General Catalyst’s Customer Value Fund (CVF), a non-dilutive credit facility underwritten against forward-looking SaaS cohorts rather than the company’s balance sheet.

Tandon framed it as a return to first principles: “Credit is not a unique concept… The reason startups shy away from it is it often puts this massive liability on your balance sheet, and if the worst case Armageddon happens, your whole company could implode.” CVF, he argued, sidesteps that risk by tying repayment to customer performance. His operating philosophy follows from there: “You actually don’t wanna use balance sheet for go-to-market. You wanna use balance sheet for R&D & these more durable long-term investments.”

Careful about the financing of the business & its structure, Tandon was straightforward abou his long-term financing goals: IPO.

“I think the best American businesses go public and become long, durable parts of retail and allow the public to invest in them,” he said, adding that given the company’s scale, with hundreds of millions in revenue, an IPO would “be fueled by going public as opposed to being hampered by it.”

The Product: An AI Operating System for Healthcare

Commure’s pitch is that healthcare doesn’t need another point solution. It needs an operating system.

“For Commure at its core is a software company that builds tools for providers & for healthcare administrators. This is a multi-trillion dollar problem, and even when you cohort it down, there’s a trillion dollars spent on healthcare admin every year.”

The platform spans clinical and financial workflows: ambient documentation, autonomous coding, patient intake and engagement, revenue cycle management, and back office automation, all orchestrated through what he calls Commure OS.

“We’ve built a series of fine-tuned language models & agents that automate all of those tasks and truly take a health system that might operate at 2-3% operating margin today because of all the labor it needs, and turn it into something that can operate at 20-30% operating margin, the top percentile healthcare practices & systems.”

He’s betting that consolidation is inevitable: “that’s not going to be a hundred different point solutions. That’s actually, actually gonna be one revenue engine and a series of agents that are orchestrated on a single platform.”

That conviction extends to how he sees competitors. Tandon drew a historical parallel to Grammatik, the early grammar checker eventually wiped out when Microsoft Word shipped autocorrect natively.

“There’s a lot of nonsense GPT wrapper businesses out there right now… they grew fast. They got a lot of VC interest. But they’re gonna be gone in three years, because companies like ourselves will completely disintermediate them.”

Scale Measured in Hours, Encounters, & Margins

The metrics Commure now operates at make the platform thesis concrete. The company supports 500+ healthcare organizations across thousands of sites of care, integrates with dozens of EHRs, and processes tens of billions of dollars in annual claims, with the vast majority of revenue cycle work completed without a human in the loop. Over 50 of the largest health systems in the country, including HCA and Tenet, are customers. ARR has doubled three years in a row.

Tandon translated the scale into human terms: “We support, I wanna say, 200 million patient encounters every year now.” On ambient documentation alone, the product that listens to a clinical encounter and generates the note and superbill, he said the company runs “about 50 million annualized appointments every year, saving at least 75 million hours a year for physicians… and that’s just productivity straight back into the American economy.”

The most affecting story he told was a quote from a physician using the product, surfaced publicly by General Catalyst’s Hemant Taneja: “I was about to quit. I was looking for another job because these long shifts were just ridiculous… And then I used the tool, the Premier Ambient tool paired with Revenue Cycle, and it literally finished fourteen charts instantly.” Tandon’s framing: “We have a massive labor shortage in healthcare. We cannot have trained people leave the system.”

Tokens, Margins, & Why SaaS Math Is Breaking

One of the more candid threads of the conversation was on unit economics in the LLM era, specifically how token costs are reshaping software margins.

“VCs have a good sense of what a SaaS business looks and feels like in terms of its margin structure, its payback, its CAC, its LTV,” Tandon said. “And one of the challenges is that these token businesses have essentially turned a lot of software businesses to service level margin. And if you play the game the same way you did two years ago, you can blow up, and you can burn a lot of cash in the process.”

Commure’s response has been to operate the business as a portfolio of distinct margin profiles. “We have, you know, we have parts of our business that run at pure SaaS margins in the high 80-90s. And you have other parts of the business, like full-cycle RCM, that have a labor component and a heavy token component, and those run at everywhere… anywhere from the high 60s in their implementation stage to the mid 70s in their steady state.”

That mixed structure is precisely what makes non-dilutive financing rational. If payback periods are well understood, Tandon argued, equity is the wrong instrument: “If you fundamentally understand the payback periods, then you should use CVF. You should use non-dilutive mechanisms to fund that expansion.”

AI in Practice: Accuracy, Evals, & the Cost of Hallucination

In healthcare, AI failure modes are not abstract. They’re regulatory, reputational, and clinical. Tandon was explicit about where the bar sits:

“Where you see rollouts get completely stopped or companies go to zero overnight is in cases where a model or a tool hallucinates or performs incorrectly and ends up impacting patient care, and so it ends up being maybe the most important thing.”

His operating answer is infrastructure heavy.

“The only way to truly ensure that is great evals. You need to have a massive data set of historical documentation or whatever the task might be, coding, voice calls, to then measure your model iterations on. Because, you know, when you release a new version of a model or a new version of an agent, it does some things amazing, but then it does a couple of things that are kinda weird, and there might be regression and performance deterioration on the fringes.”

He sees this as a structural advantage for larger, well-capitalized teams: smaller competitors can ship quickly but lack the engineering depth to catch silent regressions.

How do they get these AI systems into an industry with limited AI knowledge? An implementation playbook borrowed openly from Palantir.

“We have this forward deployed team of engineers that, you know, obviously stolen from Palantir, but I mean, the model works so well.
You take smart, hungry people that are early in their careers, and you throw them into the face of the problem within the hospital.”

Example: Commure’s first HCA deployment, he said, was “literally a couple engineers on the ground in a hospital with three physicians, and it was like two months of back and forth, how do we make this model work really well in a hospital setting or an ED setting, which are really hard settings to perfect documentation in.”

Healthcare Dynamics: Politicians, Operators, & the Regulatory Window

Asked about the state of American healthcare, Tandon gave one of his sharpest takes: “I would actually say that healthcare’s biggest problem today is it is run by politicians. It is not run by ruthless, pragmatic business operators.” The for-profit systems that are compounding fastest, he argued, share a common trait, aggressive CEOs at the top, while academic & nonprofit systems remain stuck in “design by committee & decide by committee.”

He was unusually positive on the current administration’s posture toward healthcare innovation, particularly on interoperability. “One of the biggest problems in healthcare over the last thirty years was an EMR could guzzle up all your data and then sit there, and then when you ask for it back, quote you some crazy fee to move one piece of data from one system to another.” He praised the administration for “funding the Department of Justice to come after folks that are engaging in information blocking or violating the Cures Act.”

His hottest take pushed back on the conventional narrative that American healthcare is broken. “The American healthcare system is the engine of innovation for the whole, whole world. The vaccines that are made are the best in the world. The drugs that are made are the best in the world.” The opportunity, in his framing, isn’t to import another country’s system. It’s to remove the work tax with software:

“LLMs are this gift that we’ve received to go nuke all of that work tax, and that in some ways is gonna be my life’s work, eliminating the work tax and making the system a trillion dollars more efficient.”

Hiring “Heat-Seeking Missiles”

With over a thousand employees across offices in Mountain View, New York, Boston, Nashville, Salt Lake City, and Bangladesh, hiring has become one of Tandon’s primary jobs. His framework went viral after Sequoia’s Alfred Lin tweeted an internal memo: Commure looks for “heat-seeking missile[s] for pain,” people who, in Tandon’s words,

“actively seek out the hairiest, gnarliest problems that customers have or exists generally in the business… and then surgically works to eliminate said problem. They have almost an addiction to seeking out sources of pain and blowing them up.”

The philosophy is rooted in a specific worldview about value creation. “Pain is the genesis of all capitalism. Like, there is problem, and then someone comes up with solution, and then there’s value creation and value capture in that process. Without pain, there is no capitalism. And as a result, I think we should all pray at the altar of pain.”

Without that disposition, he argued, fast-growth companies don’t survive: “Your go to market will never be as fast as the next guy over. Your products will never evolve as quickly because every product has pain.”

M&A Strategy

The same lens shapes Commure’s M&A strategy, which has included the Athelas merger and the acquisition of Augmedix, a public company purchased primarily for distribution.

Biggest M&A Lessons:

Pristine assets rarely offer price advantages, so good acquirers buy companies with fixable flaws
Culture is non-negotiable.

“There’s always a winning acquiring culture that becomes the culture of the combined business. And there are people in the company you’re acquiring that heavily opt into that. There’s also people that opt out, and that’s totally fine.”

The window for that reset, in his experience, is short, roughly the first month and a half post close.

Subscribe now

EXCLUSIVE: Coatue's Public Investments CIO on Anthropic, Semis, & AI's Biggest Market Shifts

Molly O’Shea — Fri, 15 May 2026 17:24:47 GMT

Follow the Gigawatts

Coatue’s Chief Investment Officer of Public Investments, Jaimin Rangwalla, joins Sourcery for an exclusive interview unpacking their latest Spring 2026 Public Markets Update amid one of the most unpredictable technology cycles of his 20-year career.

→ Listen on X, Spotify, YouTube, Apple

In our conversation, Jaimin breaks down how Coatue is thinking about the AI economy, including compute demand, hyperscaler spending, semiconductors, agents, and trillion-dollar private companies.

We cover:

OpenAI, Anthropic, & SpaceX are breaking into the world’s top 25 companies before going public
Anthropic adding $2.5 billion in ARR every single week
The tokenmaxxing economy
The rise of agents launching agents, creating what he calls a “digital population explosion” that will multiply everyone’s semiconductor & energy footprint by 1,000x
Why Coatue is "following the gigawatts"
The $12T funding engine behind the AI buildout
Sellers of shortage vs. buyers of shortage
The CPU/GPU flip reshaping compute demand
Coatue's $6T+ AI market estimate

Plus: the bottlenecks no one is solving, trillion-dollar IPOs coming, & what keeps him up at night.

Shoutout to Caryn, Joanna, Michael & the amazing team at Coatue!

Coatue’s Spring Public Markets Update Replay + Slides: https://www.coatue.com/blog/perspective/public-markets-update-2026-05-06

Chapters & breakdown below. Subscribe for more conversations with the investors and operators building the AI era.

Subscribe now

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

(00:00) Jaimin Rangwalla, CIO of Public Investments at Coatue
(00:56) Inside Coatue HQ
(02:48) Investor Update Kickoff
(04:36) Mapping the AI Stack
(06:02) Why Supply Stays Tight
(07:03) How Jaimin's Became CIO
(10:43) Private Giants vs Mag 7
(12:40) Market Breadth and Reordering
(15:24) Where AI Revenue Comes From
(17:04) Tokens and Economy
(19:43) Agents Change Everything
(21:58) OpenClaw Explained
(24:49) Memory Demand Explosion
(27:12) Architecture Shifts Ahead
(27:24) Agents Gain Memory
(27:58) CPU Demand Surge
(28:38) CPU GPU Ratio Flip
(30:21) Key Chip Players
(30:45) Intel Comeback Thesis
(31:41) Semis Go Mainstream
(33:24) Nvidia Mania and GTC
(33:59) Tracking Data Center Buildouts
(35:21) Jobs Lost and Created
(37:30) Sellers Versus Buyers
(40:54) Optical Breakouts
(41:27) Bottlenecks Everywhere
(44:48) Sentiment Versus Fundamentals
(47:10) Handling Volatility
(49:17) Finding New Leaders
(51:18) Trillion Dollar IPOs
(52:48) Risks and Disruptions
(55:00) Coatue Growth Story
(55:58) Staying Curious to Win

Brought to you by:

Turing—Turing accelerates superintelligence by helping frontier AI labs improve model capabilities and enabling enterprises to deploy end-to-end AI systems inside mission-critical workflows. Visit: turing.com/sourcery

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Merge—The leading provider of customer-facing integrations and agentic tools for frontier LLMs, Fortune 500 organizations, and B2B SaaS companies. Visit: https://merge.dev

Coatue’s Spring 2026 Public Markets Update: Inside the AI Supercycle

In this conversation, Coatue CIO Jaimin Rangwalla walks through the firm's May 2026 investor update and the rate-of-change framework shaping how Coatue invests across AI, from gigawatts to agents to the trillion-dollar private companies reshaping public markets. Slides from the presentation added below.

→ Listen on X, Spotify, YouTube, Apple

A New Tech Cycle, Defined by Rate of Change

Every era of technology eventually finds its defining metric. For the internet it was page views; for mobile it was app downloads. For AI, according to Coatue’s Chief Investment Officer of Public Investments, Jaimin Rangwalla, it’s rate of change, and the rate is unlike anything he’s seen in nearly two decades on Wall Street.

Coatue frames AI as the seventh and largest of the major tech waves, following Mainframe (1960 to 1980), PC (1980s), Networking (1990s), Desktop Internet (2000s), Mobile Internet (2010s), and Cloud/SaaS (2015 to 2020). The 2022 to 2026+ AI wave, in Coatue’s illustrative scaling (slide 19), towers over every prior one.

“AI is big and everyone can make these grand statements about how big it is.. but the most exciting part is the pace of the innovation, and you can look at that across, you know, how quickly companies have reached, you know, $10 billion, 30 billion, $50 billion of ARR.”

That pace is also visible in market dispersion. Coatue’s analysis of the top 1,000 global companies (slide 6) shows the spread between the 80th percentile and 20th percentile excess returns has reached roughly 85%, with the top quintile at +51% and the bottom at –33%, rivaling the 2009 and 2021 peaks of 57% and 62%.

Inside TMT, the dispersion is even more extreme. The same spread analysis applied only to the top 1,000 global TMT companies (slide 7) shows a ~180% gap, with the top quintile delivering +136% while the bottom delivers –43%. That’s nearly triple the dispersion of the 2009 financial-crisis recovery and almost double the 2021 spike. The market is not rewarding “tech” broadly; it is rewarding a very specific slice of tech and punishing the rest.

What makes this cycle unusual is the absence of the typical flattening pattern. Adoption curves at the consumer, enterprise, and revenue levels are still bending upward rather than rolling over. “Even the biggest bulls I think aren’t capturing the size of the markets,” Jaimin said. Coatue is no longer organizing coverage by traditional sector buckets. Instead, analysts are sliced across the AI supply chain itself: “We’re almost thinking what slice of the AI supply chain, and then who are the people we’re gonna have covering those slices.”

Pictured: Michael Barton, Sector Head at Coatue, Coatue’s Tech Museum

Follow the Gigawatts

Two or three years ago, Coatue’s internal mantra was follow the GPU. Betting on NVIDIA early was right, but the team knew one stock was never the whole story.

“We bet on Nvidia early, but that’s usually not enough. Like what we really characterize a big theme being that there’s many ideas across a theme because there’s so many companies that are gonna be benefiting and impacting the, the forward kind of future of a theme.”

The framework has since evolved. The new shorthand is follow the gigawatts.

“The gigawatt is almost the, the atomic unit of where the growth in AI is coming from, and it’s one of the biggest shortages as well that’s out there.”

That shift is visible in trailing one-year S&P 500 sector performance (slide 9): Semis & Hardware lead at +68% and Energy comes in at +18%, while every other sector posts negative relative returns. Industrials are at –8%, Utilities –13%, Consumer Discretionary –16%, Financials –19%, Health Care –22%, Internet –26%, Consumer Staples –30%, and Software at –40%. Coatue captions the chart simply: “Market has favored ‘sellers of shortage.’”

From the gigawatt as the atomic unit, Coatue maps outward in every direction. What are the inputs to a gigawatt? What reduces its lead time? Who are the buyers? Who consumes the tokens produced by the gigawatt’s output? Each question opens a new investable surface area.

The biggest surprise, Jaimin said, has been how long the tightness has lasted. Memory suppliers are now signing guaranteed commitments through 2029 and 2030. “The fact that it is tight, and it almost seems like it’s getting tighter every week that goes by.” The buyers, he notes, are the largest companies in the world: “They are not investing hundreds and hundreds of billions of dollars growing at these high rates into something that they think is just gonna kinda peak and trough.”

Trillion-Dollar Private Club & a $6T+ TAM

For most of modern market history, the largest technology companies became large after going public. Meta, the biggest of the Mag 7 at IPO, listed in 2012 at roughly $100 billion. That pattern has now broken. “Today you look, OpenAI’s most recent round was 800 plus billion dollars. SpaceX is most recent, when they did the transaction with xAI was $1.25 trillion. Anthropic’s last round was high $300 billions.” Three private companies have lodged themselves inside the world’s top 25 by valuation, before issuing a single public share.

The token data underwrites it. Coatue’s estimate of annualized API tokens generated by OpenAI, Anthropic, and Gemini (slide 27) sat near zero in early 2024, hit 16.4 trillion by late 2025, and is now tracking toward 35 trillion-plus on a trajectory that Coatue labels “hyper-exponential.”

That growth gives Coatue confidence in a market sized at roughly $4 trillion-plus for digital AI alone, as detailed (slide 28):

$0.3T from consumer (a billion paid users at $20 per month equating to ~30% of global subscriptions)
$2T from coding and developers (40% of the $5T global engineering and IT payroll)
$2T from non-developer white-collar work (50% of the $4T global SG&A payroll)

Add physical AI (humanoids, industrial automation, autonomous vehicles), and the TAM climbs to $6T+. Coatue’s annotation: “Could be much bigger!”

AI Revenue Acceleration & the $12T Funding Engine

If the TAM defines the ceiling, two charts in the deck show how quickly the industry is climbing toward it. Slide 30 plots the combined ARR of OpenAI & Anthropic against FY2025 revenues of established software franchises, & the line is near vertical.

The combined number sat at $3.4B in early 2024 (smaller than Datadog), reached $8.4B by January 2025 (Workday-scale), $13.3B mid-2025 (ServiceNow-scale), $23.8B by late 2025 (Adobe-scale), $37.9B by January 2026 (Salesforce-scale), and then.. OpenAI & Anthropic reached a combined ARR of ~$55B by April 2026.

Jaimin’s framing in conversation underscores the velocity. “They’re adding $10 billion plus or minus a month, almost $2.5 billion a week. Most of the companies in the SaaS universe don’t even have $2.5 billion of ARR annually, right? They’re adding that in a week.” The companies on the comparison line each took 15 to 25 years to reach their current scale. OpenAI and Anthropic compressed the same trajectory into roughly 24 months.

The question that naturally follows is how this gets paid for. Slide 29 lays out Coatue’s AI capex funding model for 2026 to 2031. Hyperscaler EBITDA contributes a base of ~$1T per year, growing at 10% annually, which compounds to $6T cumulative over the period. Layering on 3x debt-to-EBITDA leverage capacity adds another $4T. Sovereign funds, private credit, and Neoclouds contribute an additional $2T. The total: $12T of cumulative funding sources for AI infrastructure through 2031.

That number reframes the entire debate about whether AI capex is sustainable. The buyers of shortage are not stretching to fund this build-out; they are funding it from operating cash flow and modest leverage, with sovereign and private capital filling the remainder. As Jaimin put it: “They are not investing hundreds and hundreds of billions of dollars growing at these high rates into something that they think is just gonna kinda peak and trough.”

Tokens, Agents, & the Digital Population Boom

If gigawatts are the atomic unit of supply, tokens are the atomic unit of demand. “Basically, a token is the unit of thought by any AI model,” Jaimin explained. “It’s like a unit of intelligence.” Unlike human cognition, AI cognition can be priced, metered, and forecasted, and that property is reshaping how Coatue models every company in the AI stack.

The unlock that turned tokens into an economy was agents. Coatue maps the catalyst timeline (slide 20): 2015 OpenAI founded, 2017 to 2021 transformer paper and GPT-1, 2, 3 (”Transformers unlocked scale”), 2022 ChatGPT (”changed the interface”), 2023 to 2024 the NVIDIA moment as it beat Q2 2023 estimates by ~30% (”Investment cycle begins”), and 2025+ what Coatue calls the “Claude Code Moment,” the “Agentic Big Bang.”

Slide 21 captures the shift: agents launching agents removes the human bottleneck, parallelizes work by 10 to 100x, and stretches workflows from minutes to days. The model moves from “1 analyst helping you” to “1,000 analysts working 24/7.”

Jaimin describes his own workflow: open a Claude session, assign a task, open another, assign another. “You’re able to launch agents, and then you’re able to launch multiple pools of agents that also launch multiple agents below. And so to me, it’s just the sheer power of exponential growth of the number of agents and the amount of work that’s being done.” Coatue captures the consumer endpoint (slide 25) in one phrase: “Your phone becomes your remote!” A WhatsApp interface fans out into a tree of dozens of agents, each spawning subagents, each acting on the user’s behalf in parallel.

The breakout is already visible in developer tooling. Coatue’s GitHub Star History chart (slide 26) shows OpenClaw accumulating roughly 190,000 stars in its first ~4 months, a trajectory it labels “one of the fastest-scaling repositories on GitHub,” running near-vertical against React (which took ~100 months to reach ~215,000) and Linux (~80,000 over the same window). Project that forward and the implications for infrastructure could go even furhter: “In a year or two, we’re gonna have a claw that’s running hundreds of agents at all time.”

The Great CPU-GPU Flip

In the Training Era, the recipe for AI was simple: model weights plus data plus GPU. In the Agentic Era, the architecture has fragmented. Coatue’s diagram (slide 22) shows the agent at the center, surrounded by RAG-based file access, context-window memory, MCP tools, orchestration via sub-agents, the underlying LLM foundation model, and a mix of accelerators (GPUs, TPUs, and ASICs). With a clear message: “Now, memory and compute are the new bottlenecks.”

For most of the last four years, the compute ratio was lopsided: one CPU for every eight GPUs, sometimes one for sixteen. CPUs handled the final orchestration step; GPUs did everything else. Agents are quietly inverting that. “Now the ratio is actually moving from to one CPU to four GPUs, and so it’s improved kind of by 2X already, and we think it actually has a chance to flip the opposite direction, which is one GPU to four CPUs.” Some on Coatue’s team think it could go to one GPU per eight CPUs, embedding a 16x expansion of the CPU market opportunity inside the same workloads driving GPU demand.

The reason is architectural, illustrated (slide 24). Chat workloads are GPU-intensive: a single question, a single token-generating response. Agent workloads are CPU-intensive: search the web, look up a document, analyze data, update a sheet, send an email, long chains of serial subtasks, each calling tools, each requiring orchestration. Coatue’s visual flips the compute footprint accordingly: where a chat workload uses four GPUs and one supporting CPU, an agent workload uses four CPUs and one GPU.

That shift has implications for who wins. “On the CPU side, it’s Intel, AMD, and Arm,” Jaimin said. He’s particularly constructive on Intel, a company that spent years as semiconductors’ forgotten child but now has new leadership and clean math. “Some of the best thesis are the simple thesis, and you don’t wanna overthink it.”

Sellers vs. Buyers of Shortage

The single most useful framework in Coatue’s deck is the distinction between sellers of shortage and buyers of shortage. The deck’s most evocative visual is a literal tug-of-war: two strongmen pulling on a rope labeled TOKENS, the SELLER side branded with Micron, SK Hynix, SanDisk, Lumentum, GE Vernova, and Intel; the BUYER side with Amazon, Meta, Microsoft, and Google.

The economics behind the framework are extreme. Coatue’s illustrative unit economics (slide 14): a Pre-AI seller of shortage earns $40 of EBIT on $100 of revenue at a 40% margin. With a 2x price increase and only modest opex creep, today the same unit generates $135 of EBIT at a 68% margin, a 3x jump in profits. With a 4x price increase, the future case delivers $330 of EBIT at an 83% margin, an 8x jump.

“When price is the main lever of your revenue growth and you have fixed costs, your operating profit actually go up multiples of what your price increases or your revenue’s growing at.”

Real-world data (slide 15) confirms it: Micron’s operating margin has gone from a 16% five-year average to 69% today (+4x), and Seagate’s from 17% to 38% (+2x).

The market is rewarding the sellers and punishing the buyers.

According to slide 16, year to date, sellers of shortage (Broadcom, Coherent, Marvell, GE Vernova, Lumentum, Samsung, Micron, NVIDIA, TSMC, SK Hynix, Western Digital, Seagate, SanDisk) have returned ~107% on average.

Buyers (Amazon, Google, Meta, Oracle, Microsoft) have returned just ~4%. The cash-flow plumbing tells the same story.

Meanwhile, the four hyperscalers will spend a combined $680B in capex on 2026E consensus (Microsoft $157B, Google $186B, Meta $138B, Amazon $200B) against just $66B of total free cash flow ($39B, $29B, $5B, –$7B).

On the semi side? The cohort generates $525B of FCF combined: NVIDIA $184B, Samsung $132B, SK Hynix $110B, Micron $54B, TSMC $46B. Cash flow, in Coatue’s words, is “transferring from hyperscalers to AI infrastructure.”

The multiple compression on buyers is the second-order effect, shown on slide 18. Since January 2025, Meta’s forward P/E has dropped from 23x to 17x (–6x), Amazon’s from 31x to 26x (–5x), and Microsoft’s from 30x to 22x (–8x). “If memory pricing goes up 100% and Microsoft has to spend two units instead of one, they’re not actually getting two units of benefit.” Two names sit in a unique middle position: “I actually think Amazon and Google are in an un-unusual camp, and those are the two companies that we really like because they are also a bit of a hybrid in that Google has TPUs that they sell, and Amazon has talked about selling Trainium.” By owning silicon as well as deploying it, they capture some of the shortage premium themselves.

Risks, Volatility, & What Comes Next

For all the bullishness, Jaimin is a risk manager first. The biggest risk he watches is another DeepSeek moment, a breakthrough that materially reduces the power, compute, or memory required to produce the same intelligence. In the short run, such a moment compresses the sellers of shortage. In the long run, it likely accelerates adoption and expands the pie.

The other defining feature of this cycle is volatility, and Coatue’s data quantifies it. Slide 10 shows the MSCI USA Momentum Index has logged 70 days year to date (annualized) with moves of +/– 1.5% or more, the most since COVID. That compares to 71 in 2020, 47 to 48 in 2024 to 2025, 42 in 2022, and just 2 to 12 in most years between 2010 and 2019. Coatue’s annotation: “Most volatility since COVID.” Jaimin describes the qualitative experience: “The volatility we’re facing now is, wow, I’m still really bullish. AI is still doing great things, but wow, some days, some of our stocks are just down 5% or 10% on for no reason.” The discipline in those moments is to re-ground in fundamentals, and to use hedges, puts, and shorts where appropriate.

Even the Mag 7 is being reordered. Coatue’s chart (slide 8) of selected Mag 7 returns since the October 30, 2025 prior market peak shows extreme dispersion: Google has broken out to roughly +40%, Amazon is up around +20%, the NASDAQ Composite is up modestly, and the rest of the Mag 7 sits flat to down, with one name down more than 20%. As Philippe Laffont said on the call, “even within the Mag 7, the reordering of the pecking order is happening quickly.”

The underlying fundamentals remain strong. S&P 500 earnings growth is tracking at ~15% accelerating to 18% through 2026. The market is up roughly 7 to 8% year to date against 16% earnings growth, meaning multiples are actually compressing as prices rise. “At the end of the day, like, fundamentals do matter more so than sentiment, and I think the fundamentals are really strong.”

Asked what he’s most looking forward to, Jaimin returned to the firm itself. He joined Coatue in 2007 when it managed $700 million with twelve people; today it manages ~$80 billion with 200 people.

“I feel like we’re, as an organization, are at our second inflection. Maybe, probably not second, like fifth inflection. But we’re at another inflection where I think we’re really on this journey to, to really just break out.”

In a market where the rate of change is the only thing that matters, that mindset, what is changing, and how do I catch it?, may be Coatue’s most durable edge.

→ Listen on X, Spotify, YouTube, Apple

Subscribe now

BREAKING: How to Sell to the CIA

Molly O’Shea — Tue, 12 May 2026 14:52:26 GMT

Teresa Carlson, Founding CEO of the General Catalyst Institute, joins Sourcery to share the inside story of how she built AWS’s public sector business from zero revenue and two people into a $10B global powerhouse — including the legendary CIA contract that legitimized cloud computing for the entire world.

→ Listen on X, Spotify, YouTube, Apple

In this conversation, Teresa pulls back the curtain on what it was really like working alongside Jeff Bezos and Andy Jassy at Amazon, the brutal “no PowerPoint” writing culture, and the one-way vs. two-way door framework that shaped how she made decisions. She also draws sharp parallels between the early cloud era and today’s AI explosion — explaining why she thinks AI adoption is moving faster than any technology she’s ever seen, why the U.S. urgently needs an AI framework, and what startups must do to win in Washington DC.

We also get into the General Catalyst Institute’s mission to give 1,000+ portfolio companies an “unfair advantage” with policymakers, the rise of commercially-oriented government under the current administration, and why founders need to ditch the flip-flops if they want to be taken seriously on Capitol Hill.

A must-watch for founders, operators, investors, and anyone interested in the intersection of AI, policy, and the future of American competitiveness.

Subscribe now

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

(00:00) Teresa Carlson, Founding CEO at GC Institute
(01:08) The historic Kentucky Derby win
(02:38) $10B AWS blueprint
(04:56) Converting the skeptics
(12:03) Contract that changed the trajectory of AWS
(17:05) Building a world-class team
(24:01) The scaling mantra
(29:12) Lessons from Bezos & Andy Jassy
(35:33) Why is AI scaling scaring governments?
(40:23) Surviving the AI debt trap
(44:56) Why do government deals matter?
(48:55) Founding the GC Institute
(54:49) Founders storming Capitol Hill
(59:53) The unspoken rule of DC
(01:05:53) The next big AI shockwave

Brought to you by:

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Merge—The leading provider of customer-facing integrations and agentic tools for frontier LLMs, Fortune 500 organizations, and B2B SaaS companies. Visit: https://merge.dev

How AWS Sold Cloud to the CIA:

Teresa Carlson on the Deal That Defined an Industry

When Teresa Carlson joined Amazon Web Services in 2010, the company had no public sector revenue, no government contracting mechanism, and no agreed-upon definition of “cloud computing” inside the federal government. Ten years later, she had built the business into a roughly $10 billion global operation and helped land one of the most consequential enterprise technology contracts of the decade: the Central Intelligence Agency’s cloud deal.

Now Founding CEO of the General Catalyst Institute, Carlson walked Sourcery through how that business was built, what she learned from Jeff Bezos and Andy Jassy, and why she believes AI is now entering a similar inflection point with far less regulatory scaffolding in place.

Building a Market Before Selling a Product

When Carlson arrived at AWS from Microsoft, where she had run U.S. federal, the cloud category effectively did not exist in government. There was no procurement path, no compliance framework, and limited understanding among policymakers of what cloud computing was.

“We really did create the industry for cloud computing at AWS for governments around the world.”

Her early team consisted of two people and no revenue. The go-to-market strategy was deliberately grassroots. Carlson and her solution architects ran what she called “pizza or cookie evenings” in a small office in Herndon, Virginia, inviting CIOs, CTOs, and technical directors from across government to bring their laptops and try the product.

The educational gap extended to Capitol Hill. Carlson recalls early meetings with congressional leaders who assumed Amazon was there to discuss retail or taxes:

“Oh, Teresa, you’re at Amazon. You know, are you here to talk about books or taxes?”

Two structural problems had to be solved before AWS could sell at scale:

cloud needed a formal definition
it needed a compliance regime

Carlson’s team worked with NIST to establish the U.S. government’s definition of cloud computing, a definition later adopted by other governments globally, and with GSA to create FedRAMP, the first security and compliance model for cloud. FedRAMP has since become a reference standard used by regulated industries including healthcare and financial services.

The CIA Contract

The 2013 CIA cloud contract is widely viewed as the moment cloud computing became credible for regulated enterprises. AWS competed against incumbents including IBM and Microsoft, both of which fielded substantially larger bid teams.

“I had a team of four people, that was it, and we did one outside contractor that helped us. And other teams, IBM, Microsoft, others that bid it, had these huge teams bidding this thing, and they kept saying, ‘You guys can’t win it.’”

Winning required Amazon to transform internally, clearing personnel, building Sensitive Compartmented Information Facilities (SCIFs), and adapting to classified contracting environments. Carlson credits Bezos, Jassy, and the board with backing the investment despite the operational complexity.

The downstream effect on the broader market was significant:

“It was kind of the shot heard round the world because commercial enterprises were like, ‘What? If the CIA can use cloud, we can use cloud.’”

Carlson points to this as a case study in why early lighthouse customers in regulated industries matter disproportionately for category-defining technology.

Inside Amazon: The Bezos & Jassy Playbook

Carlson describes the cultural shift from Microsoft to Amazon as significant. The most immediate adjustment was the absence of PowerPoint. Amazon meetings revolved around written six-page narratives, and her first attempt to present slides to Andy Jassy and his team did not go well.

“I had a beautiful PowerPoint. Nobody would even look at it.”

The writing culture forced a different kind of preparation. Documents were expected to be precise, outcome-oriented, and explicit about risk. Carlson notes that hiding or downplaying risk was treated as a credibility issue rather than a polished pitch.

Decisions were also categorized using Bezos’s well-known one-way vs. two-way door framework:

“A one-way door is if you made that decision, you couldn’t get back, and it was, it could be amazing or detrimental. A two-way door, you could take it and you could come back.”

Another communication norm: yes-or-no questions received yes-or-no answers. Explanations followed only if requested. Carlson eventually coached her own teams on this convention, framing it as a function of organizational velocity rather than abruptness.

Hiring for a Consumption-Based Business

The cloud business model also required a different hiring profile than traditional enterprise software. Because AWS customers paid only for what they used, account executives could not rely on transactional, license-based selling. Compensation was tied to actual usage, which meant sellers had to drive genuine adoption.

“They had to delight the customer. They had to love that customer’s mission and had to go in, because they wouldn’t get anything on top of their salary until the customer started using the tool.”

Carlson built a team oriented around consultative engagement, mission alignment, and technical fluency. Solution architects were hired not only for technical depth but for the ability to listen, whiteboard with customers, and translate agency missions into workloads.

She also invested heavily in business operations and capacity planning, because forecasting in a consumption-based model required tight coordination with Amazon’s infrastructure teams in Seattle.

Cloud to AI: A Faster, Less Structured Cycle

Carlson sees clear parallels between the early cloud era and the current AI cycle, but also notable differences. The most significant, in her view, is the absence of definitional and regulatory infrastructure.

“They still really have not defined AI.
If you go look at what I said they defined cloud, they’ve not really defined what AI is.”

Adoption in regulated industries and government is moving faster than it did with cloud, often without the compliance frameworks that eventually shaped cloud procurement. Carlson views this as both an opportunity and a risk. She has used her platform at the General Catalyst Institute to advocate for a federal AI framework, arguing that a patchwork of state-level regulations is particularly burdensome for startups.

On pace, she offers a direct comparison:

“I used to tell my team at AWS, ‘Strap in, because you won’t see anything like this.’ And with AI, I’m like, ‘This is a double or a triple strap in.’”

She also expects an “optimization point” in AI spending, similar to what occurred in cloud, where customers initially expand consumption broadly and then rationalize toward the workloads that matter most.

The General Catalyst Institute

The General Catalyst Institute, which Carlson launched in September 2024 with Hemant Taneja, sits inside General Catalyst and works on behalf of the firm’s portfolio, roughly 1,000 companies, many in regulated sectors including healthcare, defense, financial services, energy, and manufacturing.

The Institute’s function is policy education and advocacy at the industry level rather than on behalf of any single company. Carlson and her team bring founders to Washington, Brussels, London, and other capitals to meet with lawmakers, agency leaders, and regulators.

The focus is on helping founders understand how to engage with government, how to position their technology in policy conversations, and how to build a go-to-market motion in regulated environments.

“Our philosophy is we want to be solution providers, not problem identifiers.”

She cites a range of portfolio companies engaging directly with policymakers, including Hippocratic AI (agentic patient follow-up), Cityblock Health (outcomes-based care for Medicare and Medicaid populations), Paragon Health (clinical trial access), and robotics companies Cobot and Standard Bots.

What Founders Need to Know About Washington DC Now

Carlson argues that the current environment in Washington is unusually open to startups, particularly compared to earlier administrations. She points to expanded contracting pathways, executive orders aimed at procurement modernization, and direct engagement from agencies including the Department of Defense.

“Under this administration, one of the things that they are doing that I’ve not seen in my entire career is how much more commercially oriented they are.”

Her practical guidance for founders engaging with government:

Treat customer references as a primary asset. Legitimacy for AWS Public Sector came from named customers willing to speak publicly about results.
Lead with the mission, not the technology. Government buyers respond to outcome-oriented conversations, not feature lists.
Show up prepared and appropriately dressed. Carlson notes that flip-flops and casual dress signal a lack of seriousness, particularly for founders unknown to the agency.
Engage early on policy. Larger companies have the resources to navigate regulation; startups generally do not, which is the gap the General Catalyst Institute aims to fill.

Looking ahead, Carlson’s priorities for the year include broader AI adoption, a federal AI framework, and permitting reform around critical minerals, energy, & data center infrastructure, areas she views as foundational to U.S. competitiveness with China.

Subscribe now

BREAKING: Uber CEO Dara Khosrowshahi

Molly O’Shea — Fri, 08 May 2026 14:01:52 GMT

Go Anywhere, Get Anything

Uber (NYSE: UBER) CEO Dara Khosrowshahi joins Sourcery for a walkthrough of Uber’s newest product launches, the future of transportation, AI, delivery, travel, & autonomous vehicles.

→ Listen on X, Spotify, YouTube, Apple

At Uber’s Go-Get showcase in New York City, Dara breaks down how Uber is evolving from a rideshare company into a global “go anywhere, get anything” platform. We discuss Uber’s new Expedia hotel partnership, AI-powered shopping agents, Uber’s autonomous vehicle strategy with Waymo, Zoox and Joby Aviation, and how AI is fundamentally changing the company’s operations and engineering culture.

Dara also reflects on leading Uber through one of the largest corporate turnarounds in tech history, scaling the company from $4.5 billion in losses to nearly $10B in free cash flow.

Plus: how to actually raise your Uber rating.

Subscribe for more conversations with the CEOs, founders, and investors shaping the future of technology, AI, mobility, and infrastructure.

Subscribe now

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

(00:00) Dara Khosrowshahi, CEO at Uber
(00:48) The biggest bet of 2026
(02:04) The accidental feature
(04:16) Your next Uber might come with a Coffee
(05:54) Uber One’s global takeover
(06:32) Growth hack to outpace rivals
(08:04) Replacing overpriced room service
(11:25) The Expedia reunion
(14:10) The 10-minute airport hack
(17:05) The $10 billion comeback
(19:00) Barry Diller’s golden advice
(22:49) Uber’s internal AI shift
(26:59) Will AI make Uber apps obsolete?
(27:36) Self-driving cars
(28:47) AI’s biggest misconception
(29:39) Hack your Uber rating
(31:36) Junior Devs coding for millions

Brought to you by:

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Merge—The leading provider of customer-facing integrations and agentic tools for frontier LLMs, Fortune 500 organizations, and B2B SaaS companies. Visit: https://merge.dev

From Expedia to Uber and Back Again: Dara Khosrowshahi

At Uber’s GoGet 2026 showcase in downtown New York, Dara Khosrowshahi, CEO of Uber (NYSE: UBER), walked through a product lineup that included one announcement with personal weight behind it: a partnership with Expedia, the company he ran for twelve years before taking the Uber job in 2017.

I met with Dara after the event to walk through the announcements and talk about the philosophy behind Uber’s product expansion, the lessons that shaped him as an operator, and his vision for an AI and AV powered future.

→ Listen on X, Spotify, YouTube, Apple

A Full Circle Moment With Expedia

Uber (NYSE: UBER) started as a rides app, expanded into Eats, and is now staking a claim in travel, with hotels rolling out through a partnership with Dara’s former company.

“We sure did. So we announced actually now making hotels available on Uber.
You know, we started as the Go App, just rides obviously. And then we went with Uber Eats and it was just an idea.
Now Uber Eats is just as big as their rides business. Now we’re going to travel.”

The Expedia tie up wasn’t something Dara could touch directly. Still on Expedia’s board, he had to recuse himself from negotiations entirely.

“I had to recuse myself from the process because really I’m still on Expedia board. So from a governance standpoint, it’s like, ‘Hey, Dara, you can’t.’”

The strategic logic was straightforward. Travel had always been a major use case on Uber, and hotels were a natural extension of a platform already wiring up everything that moves.

BIG Numbers: FY 2025 & Q1 2026

The product expansion is happening on top of one of the strongest financial setups in tech. Uber (NYSE: UBER), publicly listed on the New York Stock Exchange, closed out 2025 with five consecutive years of 20%+ growth and rolled into Q1 2026 with momentum that exceeded that pace.

Full Year 2025

Trips: 13.6 billion, up 20% YoY
Gross Bookings: $193.5 billion, up 19% YoY (20% constant currency)
Revenue: $52.0 billion, up 18% YoY
GAAP Income from Operations: $5.6 billion, up 99% YoY
GAAP Net Income: $10.1 billion
Non-GAAP Operating Income: $6.5 billion, up 50% YoY
Non-GAAP EPS: $2.45, up 35% YoY
Free Cash Flow: $9.8 billion, up 42% YoY

Q1 2026

Trips: 3.6 billion, up 20% YoY
Gross Bookings: $53.7 billion, up 25% YoY (21% constant currency)
Revenue: $13.2 billion, up 14% YoY
GAAP Income from Operations: $1.9 billion, up 57% YoY
Non-GAAP Operating Income: $1.9 billion, up 42% YoY
Non-GAAP EPS: $0.72, up 44% YoY
Free Cash Flow: $2.3 billion
Uber One Members: 50 million, now driving half of all Gross Bookings across Mobility and Delivery

CFO Balaji Krishnamurthy put it on the Q1 call:

“From this position of strength, we’re investing with conviction in the significant opportunities ahead, while taking a capital-efficient approach to AVs and embracing AI to drive growth and productivity.”

Inconic Turnaround: From $4.5B in Losses to ~$10B in Free Cash Flow

When Dara took over Uber (NYSE: UBER) in 2017, the company was bleeding cash and battling a culture crisis. By the end of 2025, it was generating $9.8 billion in free cash flow.

“A lot of people talk about Uber. You know, when I came there, obviously there was a lot of controversy in terms of the culture, et cetera. But the team had built a good business in terms of demand and the brand and our presence globally. So while things were not easy, I kind of was standing on the shoulders of giants.”

The strategy came down to supply, safety, and platform leverage. Being multi product turned out to be a structural advantage competitors couldn’t easily replicate.

“We compete with lots of people, but most of our competitors are monoline.
They either have rides or they have eats.
The fact that we are a complete platform allows us to grow faster than our competitors and be more profitable than our competitors as well.”

That platform thesis is so durable that Dara wouldn’t be surprised if Eats overtakes Rides within a decade, especially as it expands beyond food into grocery and every other local business category. The Q1 numbers back that up. Delivery Gross Bookings grew 28% YoY to $26.0 billion, basically tied with Mobility’s $26.4 billion, and Delivery segment operating income jumped 43% YoY.

Lessons From Barry Diller

Long before Uber, Dara was an analyst at an investment bank when Barry Diller pulled him into the IAC orbit. Two lessons from Diller still anchor how he leads.

The first is being willing to go against the grain.

“Barry was the ultimate counter puncher. He kind of took on ABC, NBC, CBS… and he built Fox, which was a new product. I’ve always been comfortable going against the grain, and it’s one of the things that encouraged me to join Uber. From the outside, things looked really, really difficult.”

The second is being a relentless truth seeker, even when the org chart makes that hard.

“As you move up in organizations, while you get a broader view of the company, a lot of times you actually don’t understand what’s really going on in the company. And so you do have to be that truth seeker and it requires you not to manage in a way that you might be taught.”

In practice, that means skipping the decks and going several layers deep into the organization.

“I go and I sit down with engineers and product members two, three, four levels down.
We run no decks, really talk about what’s going on. And I’m very, very honest with them in terms of what our challenges are. I don’t talk corporate-speak to them.
Everyone has a pretty good BS meter. And if a senior member of the team is BS-ing his team, then they’re gonna BS you right back.”

Why Uber Was Built for the AI Era

Dara made the case that most companies are wired for certainty, and Uber (NYSE: UBER) never had that luxury. That turns out to be a major advantage now that the entire tech industry is moving toward probabilistic systems.

“Most companies want certainty, but in real life things can go wrong. You can order that Uber and we told you it’s a four-minute ETA, the driver cancels.
We have to actually get you another driver. So there’s all kinds of uncertainty that happens in the real world, and we have been built on top of that.
So everything that we built has been algorithmic in nature.”

Internally, AI is reshaping how Uber ships product. Engineers are using Claude and Codex, agents handle code reviews and debugging, and customer service is being rebuilt from first principles, training agents on outcomes rather than rigid policy trees.

“Now with our agents, we can train them on what’s my goal? My goal is to make Molly, who’s an Uber One member, a happy customer. And so the agents themselves can now reason around the outcome that you’re going after.”

But Dara is clear eyed about the public perception challenge ahead.

“The fears of job displacement are real… We need to bring out these AI products and surprise and delight consumers every day so that they see AI working for them, not replacing the work that they do.”

Shop For Me: A Feature Users Invented

One of the most distinctive launches at GoGet 2026 wasn’t dreamed up in a product offsite. It came from watching what users were already doing.

“This was actually inspired by our users who were hacking into essentially our system.
We have a service called Uber Courier… What we saw is people who are using Courier were hacking the system and they were asking couriers to go to a store for them and buy something for them.”

The result is Shop For Me. Upload a picture, describe what you want, and a vetted top rated shopper picks it up from any store, with approval and payment handled in app. It’s a clean illustration of Uber’s bias toward observation over assumption.

Coffee in Your Uber Reserve

Cross platform engagement is a core thesis at Uber. Riders who also use Eats spend roughly 3x more and are far more loyal. With Uber One members now driving half of all Gross Bookings, deepening that overlap is a real growth lever. Coffee delivery into Uber Reserve and Uber Elite rides is a small but pointed example of how the company is bridging its two main businesses.

“We have users who use Uber rides and we have people who use Uber Eats and we’re trying to create cross-platform interaction.
Consumers who use both services usually spend three times more. They’re much more loyal to the platform. So we want to introduce the different services in a delightful way.”

It’s a low friction way to nudge a rides only user into experiencing delivery without making it feel like an upsell.

Uber One Goes Global

Uber One has grown to 50 million members, and one of the most consistent pieces of feedback was that benefits weren’t expanding globally. That changes now.

“People who use Uber tend to travel a lot… One of the additional benefits that they were asking for is, ‘I want to get Uber One credits and benefits when I’m traveling, not just when I’m at home.’ So of course we listen to our most valuable users.”

International credits and member discounts are now part of the package, deepening the value of the subscription for the travelers who use Uber most.

Uber Eats Travel Mode

Anyone who has stayed in a hotel recently knows room service has quietly fallen apart. Uber’s answer is to slot Eats into the gap with a dedicated travel mode.

“Room service at hotels has suffered over the past 10 years.
You never know if it’s gonna be available. You often have to wait for a long time. It’s often really expensive.”

When Uber detects you’re in a hotel, the app shifts into travel mode, opening up the full local restaurant inventory and adjusting delivery logistics for hotel pickup, whether that’s the lobby or your door.

Hotels on Uber

The headline announcement of the day. Powered by Expedia, Hotels on Uber launches with 700,000 properties and VRBO vacation rentals coming next. The economics are designed to pull Uber One members in.

“Uber One members, when you book hotels on Uber, you get 10% back on all of your bookings, regardless of which hotels you book. And on 10,000 hotels kind of a rolling list, you get at least 20% off.”

Combined with Uber’s existing payment and identity infrastructure, the booking experience compresses into something closer to ordering a ride than navigating a traditional travel site.

The AV Platform Strategy

Uber’s autonomous vehicle approach mirrors how it built the human driver network. Every safe driver welcome, regardless of who built them.

“We believe we have a platform strategy with AVs. Just like we want every safe human driver on the platform, we want every safe robot driver on the platform, whether it’s a Waymo or a Zoox or a WeRide.”

Uber is already live with Waymo in Austin and Atlanta, with Zoox, Wave, and WeRide launches in motion.

Joby Aviation is in the mix too, with the goal of eventually offering air taxi rides directly through the Uber app.

“There are a number of companies that are going to get to the finish line in terms of developing a digital driver that’s safer than a human driver. We think it’s going to be great for society. We want to help those companies get to market.”

What’s Next

Watching Dara walk through the showcase, the throughline was clear. Uber (NYSE: UBER) is leaning hard into building, shipping fast, and testing new ideas. As he put it,

“One of our philosophies is we just wanna build stuff. And if you’re not failing… we want to move faster. We want to take risks.”

From a full circle Expedia partnership to a platform built for both human and robot drivers, Uber’s next chapter is being written at the intersection of cross platform behavior, AI, and a willingness to ship before having all the answers. With $9.8 billion in free cash flow behind it and Q1 2026 already accelerating off that base, Uber has plenty of fuel to make those bets at scale.

→ Listen on X, Spotify, YouTube, Apple

Subscribe now

BREAKING: First-Ever Full Tour of Figure's Humanoid HQ

Molly O’Shea — Fri, 01 May 2026 14:49:42 GMT

HUMANOID TOUR

For the first time ever, Figure is opening every door. Sourcery gets the first full tour of Figure's humanoid robotics campus with Founder & CEO Brett Adcock — and we see everything.

→ Listen on X, Spotify, YouTube, Apple

Figure is the first-of-its-kind AI robotics company bringing a general purpose humanoid to life, designing, building, & testing every robot in-house. Brett gives us an exclusive look through every department on their San Jose campus: BotQ Factory, Testing, Design, Demos & more.

System integration lab: where robots are stress-tested with software faults & physical pushes
Helix AI: team floor where the controls & neural network engineers train the vision-language-action model that runs onboard every Figure robot
Reinforcement learning & stability testing: where Figure demos the Vulcan project — surviving a lost knee mid-task
Home: environment where Figure 03 autonomously tidies a living room using their Helix neural network (no teleoperation)
BotQ: manufacturing facility where heads, batteries, and limbs come together on the assembly line, including the custom-built battery line & end-of-line burn-in bays
Industrial design studio: (opened publicly for the first time!!) housing every generation of Figure robot ever built, including: Figure 01 with its Frankenstein forearms, Figure 02, & the sleek Figure 03 that recently appeared at the White House, plus the evolution of Figure's hands & feet

On our walk/talk Brett shares why he believes humanoid robots may achieve AGI before any other form factor, why Figure pivoted entirely from hand-coded controls to neural networks, & teases that Figure 04 will be their "iPhone 1 moment."

This was so much fun!

Big thank you to Brett & the Figure team for opening the doors to us!

Subscribe now

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

(00:00) Inside Figure’s Robotics Campus
(00:48) The humanoid factory
(03:18) First robot guest at the White House
(05:29) Controlling a robot with infinite poses
(10:46) The truth about robot failures
(13:00) Attacking a humanoid robot
(16:12) Building a general purpose robot
(23:05) The "Never Fall" protocol
(28:56) Is the home robot teleoperated?
(33:36) Leasing a 24/7 robot
(35:01) Can a humanoid build a real car?
(43:32) From flying robots to humanoids
(45:59) The hidden path to physical AGI
(56:21) Figure's secret design studio
(01:00:44) Figure 4: The biggest leap in robotics
(01:06:25) Training robots in spandex
(01:10:26) Westworld, TIME Magazine, and Deadmau5

Brought to you by:

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Merge—The leading provider of customer-facing integrations and agentic tools for frontier LLMs, Fortune 500 organizations, and B2B SaaS companies. Visit: https://merge.dev

Inside Figure: The Humanoid Robot Company Building Toward AGI

→ Listen on X, Spotify, YouTube, Apple

Founder ^ CEO Brett Adcock guided us through every department of the company’s humanoid headquarters, from the system integration lab to the manufacturing floor to the previously off-limits design studio housing every robot Figure has ever built. What follows is the most complete public look at Figure to date.

Figure is a pure-play humanoid robotics company founded less than four years ago. The company designs, builds, and tests its robots entirely in-house, with roughly 500 employees across the company and a few hundred robots on campus at any given time.

“My goal for this building is I want more robots, humanoid robots, than humans walking around.”

Lobby & Charging Bay: Robots That Run Themselves

The tour begins in Figure’s lobby, where humanoid robots greet visitors, hold concierge-style conversations, and dock themselves on inductive charging pads built into the floor. The robots charge through their feet at two kilowatts, running for four to five hours on a charge, all without human intervention.

One robot in the lobby wears an American flag patch. It was, Brett explains, part of the first humanoid robot deployment in White House history just one week before our visit.

“We got a call asking to be basically the first humanoid group ever to put robots in the White House. So last week we basically had the first humanoid robots in history there, doing stuff, talking, greeting folks.”

The fleet operates 24/7 with no human babysitter. Robots dock at around 10 to 15% battery, and another robot rotates in off the dock to continue work, a swap that takes about 30 seconds. This continuous operation is the foundation for everything else Figure is doing.

System Integration Testing: Trust But Verify

Inside the main engineering building, dozens of robots run nonstop in test bays, doing burpees, squatting, getting up off the ground, and stress-testing every new piece of hardware or software before it ships. The banner over the lab reads “Trust But Verify,” and the team’s job is to find every possible failure mode before robots leave the building.

The mathematical complexity of controlling a humanoid is staggering. With roughly 40 motors that can each rotate 360 degrees, the number of possible body configurations is, as Brett put it, “360 to the power of 40… more body positions than atoms in the universe.” That’s why Figure abandoned traditional code-based control entirely. “You can code your way out of this. And I think that’s a full dead end. Or you can run an AI first strategy in a market. And that’s what we do here at Figure.”

Most of Figure’s failures today are no longer hardware. “Hardware’s gotten really robust. It’s basically a software issue.” The team runs durability and reliability tests every single day, validating new cameras, structural components, neural network weights, and firmware before any change reaches the broader fleet.

Helix AI Floor: Neural Networks Replace Code

Figure’s controls and AI team, which Brett calls “one of the best controls teams in the world,” works on the same floor as the test bays. Their product is Helix, Figure’s proprietary vision-language-action model that runs onboard every robot. Helix takes camera input and outputs joint-level commands for the entire body 50 to 200 times per second.

The team has fully transitioned from a model-based control stack to reinforcement learning. As Helix lead Moritz explained during the tour, the team simulates every conceivable real-world disturbance, including friction, gravity variations, and external forces, inside a physics simulator, trains the controller in that environment, and then “zero-shots” the policy onto the physical robot. I tested it firsthand by pushing the robot as hard as I could.. surprisingly, it really did feel heavy, sort of like human resistance, and absorbed each shove, staying upright the whole time.

Figure’s “Never Fall” initiative goes further. The Vulcan project allows the robot to lose a knee, hip, or ankle joint mid-task and still hobble safely to a charging station. “Months ago, we’d lose a knee, the robot would just fall. Now the robot loses a knee, it can either continue to do work or it can just hobble off, and another robot comes in and fills in.” All of this is learned inside a single neural network controller.

Home Environment: Tidying a Living Room

In a mocked-up home space, a Figure 03 robot autonomously tidies a living room, wiping down a table, putting away cups, and cleaning up toys. It’s running entirely on Helix, with no teleoperation. “There are rumors that these are teleoperated,” I noted. Brett’s response: “For sure, not teleoperated. These robots are running purely autonomously from an onboard AI policy called Helix running onboard the robot in the torso.”

The same neural network that handles housework also handles commercial use cases like logistics and manufacturing. That generality is the entire point. “The reason why humans are so great is the same humanoid can go over and do logistics and healthcare and manufacturing, or do dishes, just like we can.” Figure trains a single model and deploys it across every environment.

The home product isn’t yet for sale, but the target pricing is concrete. Brett wants to lease the robot for “somewhere like a car lease, maybe like four or five, six hundred bucks a month.” For Brett personally, the killer use cases are obvious: laundry every day, dishes every day, and tidying the house multiple times daily.

BotQ: The Manufacturing Floor

Next door, Figure manufactures every robot at a facility called BotQ, short for “robot HQ.” This is where heads, batteries, arms, legs, hands, and torsos are built and assembled. Figure designs and builds nearly all of it in-house, including the custom machine that performs cell-level testing and voltage balancing on every battery pack.

Each battery is a 2.25 kilowatt-hour pack engineered with aircraft-grade thermal runaway protection. “The requirement is you want no flame to ever exit the pack. You don’t want a robot on fire out in the world.” Figure has never had a robot catch fire. The pack is also structural, designed to absorb impact loads during a fall without compromising the cells inside.

The cost reductions across generations have been dramatic.

Figure 01 cost hundreds of thousands of dollars per unit because it was optimized purely for development speed.
Figure 03 brought that down by roughly 90%, to well under $100,000 per robot.

In March, Figure built more robots in a single month than it had in the company’s entire history combined. Eventually, Brett wants the lines fully automated: “It’ll just be full lights out manufacturing. We’ll have robots only building robots.”

BMW & the First Car Built by a Humanoid

Walking past a parked BMW X3 between buildings, Brett reveals one of Figure’s most significant commercial milestones. The car was built last year on a BMW factory body shop line where Figure deployed humanoid robots for six months. It is, to Brett’s knowledge, “the first car in the world built by a humanoid robot that we’re aware of.”

Brett bought the first four units off that line. Three sit on Figure’s campus and one is parked at his house. “It’s a collector’s item now.” The deployment served as the first major proof point that Figure’s hardware and software stack could survive a real industrial environment, with the speed, accuracy, and reliability bar set by human factory workers.

That bar is high, and Brett is candid about how hard it remains. Robots have to operate safely, almost never fault, self-correct when they do, and match human performance on speed and precision. “When they do fault, they need to understand that and self-correct. And then we need to be able to do useful human work at human performance. So that’s a hard bar to hit.” The BMW deployment was the first time Figure cleared it.

The Grid: Last Line of Defense

Adjacent to BotQ sits a separate facility called The Grid, named in homage to a sci-fi film. Brett, a self-described “total sci-fi geek” whose favorite film is Contact, uses The Grid as the last validation step before any code or hardware change reaches a customer. “You don’t want robots having any problems. We want to run them close enough to heavy operations that we would see out in the real world.”

The Grid runs robots 24/7 under simulated client-site conditions, including holidays, weekends, and 2 to 3 in the morning shifts. The goal is to expose any failure that wouldn’t surface in shorter testing cycles. By the time a robot leaves The Grid, it has been pushed harder than most customers will ever push it.

Together, The Grid, BotQ, and the System Integration Lab form a three-stage pipeline: build, validate, and stress test. Brett sees the operational discipline of running this pipeline at scale as one of Figure’s biggest competitive moats.

“We have to be very diligent about when we find issues, how to track them, how to do fault analysis really quickly, & then how to solve them across the global fleet.”

The Design Studio: Every Figure Robot Ever Built

The final stop, and one Figure has never opened publicly before, is the industrial design studio. Lined up in sequence are Figure 01, Figure 02, and Figure 03, plus prototype hands and feet from every generation. Figure 01, the first-generation robot, walked within a year of the company’s founding. “We had Figure 01 walking before we were a year old. We think it’s probably one of the fastest times in human history.”

Figure 01 was a “cyberpunk” robot with a backpack-mounted battery and the company’s first attempt at a tendon-driven hand, an architectural choice the team eventually abandoned. To pivot quickly, Brett swapped foot motors into the forearm, creating an oddly bent wrist that bent halfway up the arm.
Figure 02 moved the battery into the torso, tripled the compute, and adopted an aircraft-style exoskeleton structure where the outer skin carries the load.
Figure 03 reduced weight, slimmed the silhouette, added soft foam padding, introduced cleanly swappable fabric clothing, and integrated tactile-sensing hands.

But the robot Brett is most excited about is the one not yet on display.

“We’re like almost in flip phones. And now we’re entering like iPhone One moment. I think maybe Figure 04 will be our first iPhone One moment for this, where it’s just radically different.”

Hand Evolution: 5 Generations

In the design studio, Brett unveils something Figure has never publicly shown before: the original tendon-driven hand from Figure 01. Modeled after human biological architecture, the hand placed motors in the forearm and used tendons to drive each finger. Brett now considers it the wrong engineering choice. “We learned a lot about the hand of like, why it’s probably not the right direction. So our first generation hand is, how do we get a really dexterous hand built, which is really good for intelligence and AI. And it ended up becoming the wrong engineering choice.”

The pivot away from tendon drives created an unexpected design problem. To replace the wrist motors quickly, Brett repurposed motors from the robot’s feet, resulting in a wrist that bent halfway up the forearm. “I was so ashamed. I’m like, we’re gonna get this thing out… and I don’t think I’ve ever had a single person in three years ask why the forearm was like this.”

Figure has now iterated through five generations of hands, culminating in a recently teased high-degree-of-freedom hand with human-equivalent dexterity. Brett considers this the most critical component in the entire roadmap to AGI. “If we want to solve AGI and get to human intelligence in the physical world, it’s all going to start here with the hands for us.” The reasoning: humanoids need to passively learn from human video at scale, and the robot’s hand needs enough degrees of freedom to map every human motion at test time.

The Spandex Data Operation

Walking between buildings, I saw employees in full spandex suits around campus. Brett confirms Figure runs a large-scale motion capture operation to train Helix on human movement. “We’re trying to do joint-level tracking and different types of data collection, learning from humans. Our training set is, how do we, like, we’re a humanoid, we need to learn from humans at scale.”

This data engine is what powers Helix’s generalization. Figure runs both internal motion capture sessions and external collection in real-world environments, building a dataset of human joint movements and image-conditioned action policies. It’s also the bottleneck. “The biggest blocker for us now of going from where we’re at today to large-scale deployment is data. We need an enormous amount of data, diverse, really high quality, across the world.”

When Figure deploys robots at customer sites, the data flywheel will close. Robots will collect operational data, send it to centralized training jobs, and receive over-the-air policy updates with the latest Helix weights. Brett notes the company is committed to anonymizing user data, with the focus on robot state information rather than personally identifiable content.

How Figure Got Here So Fast

Asked how a four-year-old company built one of the most complex robots ever attempted, Brett points to his prior company Archer Aviation, where he built electric vertical takeoff aircraft. The underlying architecture, he argues, is nearly identical.

“My company before this designed flying robots at Archer, and it’s got the same properties. We have a battery pack, electric motors, control software, embedded systems, and sensors. That’s a robot.”

Archer’s aircraft are, in Brett’s words, “highly over-actuated,” with variable-pitch propellers, tilting leading-edge actuators, and moving flaps across the wings and tail, totaling 24 degrees of freedom. Figure 03 has just over 40. The systems engineering DNA carried over directly.

“When I started Figure, we had a very crisp and clear vision for the product and engineering roadmap. And we just went really hard, building a team to 40 and putting the right resources in place to design stuff really fast.”

That speed shows up in the gap between vision and execution. Figure 01 walked inside a year. Figure has now manufactured hundreds of robots, deployed them with BMW, run them autonomously in a home, and put them in the White House, all in under four years.

The Path to AGI

Closing out the tour, Brett returned to the bigger thesis behind Figure. The company isn’t building robots to replace specific jobs. It’s building them as the physical embodiment for general intelligence itself. “It actually might be the case that we get to artificial general intelligence first in these embodiments,” he said. “Real world interaction with our environments, that’s the last missing piece to get to true AGI.”

Whether or not that turns out to be right, what’s undeniable after walking the campus is the operational discipline behind it. Figure has compressed roughly a century of automotive-style hardware development into less than four years, while simultaneously running a frontier AI lab, a manufacturing operation, and a 24/7 deployment fleet. Few companies in the world are attempting all four at once.

The first full tour of Figure makes one thing clear: humanoids are no longer a research project. Figure is a real an industrial company building toward a future where humanoids do dishes, build cars, and possibly arrive at general intelligence, one trained policy at a time.

→ Listen on X, Spotify, YouTube, Apple

Subscribe now

But wait, there’s MORE..

BREAKING: Brett Adcock, CEO of Figure

Thu, 30 Apr 2026 17:59:26 GMT

HUMANOIDS ARE COMING

Brett Adcock, Founder & CEO of Figure, the $39 billion company tackling one of the most ambitious challenges in technology — building a general-purpose humanoid robot that can do real human work, both in the factory & the home.

→ Listen on X, Spotify, YouTube, Apple

About 50% of global GDP is human labor.. in this episode of Sourcery, we go inside Figure’s headquarters for a sit-down interview on the future of robotics, AI, & jobs.

Brett shares why humanoid robots are already working today, how Figure plans to scale from thousands of units this year to 1 million per year, & why he believes this could become the biggest business in the world.

Backed by nearly $2B across rounds — including high profile investment from Jeff Bezos, Microsoft, Nvidia, & Amazon — Figure recently 15x’d its valuation to $39B in just 18 months. We also cover Brett’s hot-topic decision to part ways with OpenAI, the challenges of building physical intelligence, & what it now takes to solve one of the hardest problems in engineering.

We cover:
• Why humanoid robots are finally real
• The roadmap to millions of robots
• Scaling Figure’s production & meeting demand
• Why robotics is an intelligence problem
• The OpenAI partnership — and breakup
• The risks of building physical AI (Capital, Economics, Valuation)
• Brett’s vision for general robotics

Subscribe now

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

(00:00) Brett Adcock, Founder & CEO at Figure
(01:13) Hot take on Humanoids
(04:09) Race to a Million Robots
(06:24) Building Figure From Scratch
(08:15) Leaving Archer aviation
(09:25) Post-IPO move
(12:54) Cover, Hark & NASA's IP
(16:06) How Figure Outgrew OpenAI
(19:18) Founder performance playbook
(24:24) Learning from Bezos and Jobs
(29:19) Biggest risks to the business
(31:26) Building the World's First Physical AGI

Brought to you by:

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Merge—The leading provider of customer-facing integrations and agentic tools for frontier LLMs, Fortune 500 organizations, and B2B SaaS companies. Visit: https://merge.dev

Brett Adcock: Building Physical Intelligence at Scale

Brett Adcock, Founder and CEO of Figure, is pursuing a version of robotics that has historically remained out of reach: general purpose humanoids capable of performing economically useful work across environments. The objective is not narrow automation but systems that can perceive, reason, and act with enough reliability to integrate into real workflows.

“We just want humanoid robots to work, they’re working now & it’s pretty simple.. we’re seeing robots do everyday things like clean up a living room, do commercial work. It’s cool to see that this is gonna happen in the next few years.”

The emphasis is not on demonstration, but utility. A robot that cannot sustain useful output is economically irrelevant regardless of technical sophistication.

→ Listen on X, Spotify, YouTube, Apple

Progress to Date

Figure plans to build thousands of robots this year, with parts already in house and production ramping. The company hit record production in March and aimed to triple that by May. From there, the trajectory targets tens of thousands, hundreds of thousands, and an eventual goal of 1 million units per year.

A small batch of Figure robots deployed to BMW last year and ran every day for six months. “We refactored our whole approach to how to commercialize the software and AI systems after that.” That refactor produced Helix 2, Figure’s second generation AI model.

The hardware progression reflects a similar pattern of compounding reliability. Figure 1 ran for about an hour before faulting. Figure 2 reduced fault rates to roughly once a day. Figure 3, the platform now operating across the company’s facility, sees faults across the fleet on a weekly basis rather than daily. “As we’re growing the fleet, the absolute number of faults are rising & we gotta go figure out how to solve those.”

How Much Has Brett Invested..

Brett self funded Figure before any outside capital, putting in a reported $100 million of his own money to start the company. “We got to a million a month of burn in four months. It was no joke. We had a 40 person team in four or five months. They were very good.”

Figure has since raised nearly $2 billion across all rounds, with backing from Jeff Bezos, Microsoft, Nvidia, Amazon, Brookfield, and Salesforce. The September 2025 Series C, led by Parkway Venture Capital, brought the company to a $39 billion post money valuation.

For Brett, the binding constraint isn’t capital. The challenge is solving autonomous operation at human level reliability, not financing the company through it.

The OpenAI Debacle

OpenAI’s startup fund participated in Figure’s Series B in early 2024 alongside Microsoft, Nvidia, Amazon, and Jeff Bezos. The deal came with a collaboration agreement to develop next generation AI models for humanoids. For about a year, the teams worked together on language models for robotics.

Then Brett ended it. “It got to a point where we were just like, our team internally that was designing these models were running circles around OpenAI. We were just way better at this,” he said. “And I think there was also like some interest as OpenAI was watching us getting into robotics. And so I fired him.”

Brett’s team came from robot learning backgrounds with over a decade of domain experience, which he distinguishes from large language model development. “We were better about testing on the robots, training the models, like all of it.” Ending the partnership gave Figure full ownership of Helix and full control over its development timeline.

IP, Security, & Owning the Model

Figure operates with a heightened security posture. Phones are covered on entry and areas of the facility are restricted to protect what Brett considers some of the most valuable IP in robotics.

“What we’re doing is like a very high IP risk. We really think about carefully about our engineering, CAD & software, making sure it’s very secure from a cybersecurity perspective and internal security perspective.”

The posture tightened after a specific incident. “One day we’re looking up and we’re like, it was in this office and we looked out the corner of the window at the top. There’s like a drone sitting there looking in the office.” The team tinted all the glass and tightened both physical and digital security after that. “We’re designing some crazy stuff, so we wanna protect it as all costs.”

The decision to end the OpenAI partnership reflects the same logic at the model layer. Working with an external AI lab on Figure’s core intelligence system meant sharing visibility into how the company was solving robotics, while OpenAI was beginning to show interest in the space itself. Cutting the partnership empowered Figure to develop Helix entirely in house.

Owning the full stack, from CAD files to the AI models running on the robot, is what Brett sees as the durable position.

“Without that you’re left with the mercy of some vendor. And then if that has an issue, how are you gonna go solve it? If it’s got a code problem, do you understand it? Can you QA it? Can you fix it? Can you patch it?”

Misconceptions About the Company

Brett’s central pushback is that humanoid robots get dismissed as either teleoperated or years away from working. “It’s kind of really difficult to see what’s really happening in the space without coming on site and really seeing stuff,” he said. “We just want humanoid robots to work, and they’re working now.”

A second misconception is that humanoids are a manufacturing race. Brett rejects that framing directly. “In my mind, this is not a manufacturing problem. This is an intelligence problem.” Figure designs nearly every part of the robot in house, including motors, sensors, batteries, and kinematics, but the harder problem is the AI system above the hardware.

The third misconception is that Figure can outsource its way to scale. “Without that you’re left with the mercy of some vendor. And then if that has an issue, how are you gonna go solve it? If it’s got a code problem, do you understand it? Can you QA it? Can you fix it? Can you patch it?” Vertical integration, in Brett’s view, is what allows Figure to debug, iterate, and own its trajectory.

Use Cases

Figure is targeting both commercial and home environments. On the commercial side, demand outstrips what the company can ship. “I could put so many robots into commercial customers today if they were already ready.” A larger batch is heading to multiple commercial customers this year, with announcements expected in the coming 90 days.

In the home, Brett’s bar is significantly higher. “I wanna be able to put a robot into a home and be able to do seven to 10 hours of work successfully without failures with no human intervention, and do that every day forever.” No company has demonstrated that level of autonomous reliability.

The end goal is general robotics, not task-specific automation. “A robot that can do everything a human can.. almost like a feeling of a human in a body suit that you can talk to, can look at you, reason, visual understanding. And you can drop into any place and just look around, reason and understand.”

Risks

Brett is candid about the difficulty of the problem. “The humanoid thing’s just so hard. I can’t even explain it very well. Getting the robots to do the things we showed you today.. it has almost killed me.”

The core challenges Brett lists are now more on the software side. Achieving autonomous operation without human intervention, getting AI policies to work reliably, and building physical movements that don’t fail. “We have like a fun house of problems and it’s never ending problem city for the company.”

The underlying technical risk is the open ended state space a humanoid encounters in the real world. “The amount of states the robot can be in is so high. You can’t reasonably predict what the robot’s gonna look like at every single timestamp everywhere in the world.” Cars only drive on roads. Humanoids operate everywhere humans do.

As the fleet grows, faults grow with it. “As we’re growing the fleet, the absolute number of faults are rising and we gotta go figure out how to solve those.” The most important benchmark Brett names is putting a robot into a home and having it do seven to ten hours of work successfully every day with no human intervention.

The Future of Labor with Humanoids

Brett is direct about why Figure’s $39B valuation doesn’t concern him. In his framing, humanoid labor is not just a large market. It is the largest addressable market in the AI era, and the valuation reflects a fraction of what the category becomes if the technology works.

The math starts with global labor. Roughly half (50%) of global GDP is human labor in the commercial market, which Brett pegs at $30 to $40 trillion in annual wages paid to workers worldwide. “This will build the biggest business in the world.. We will have the ability to ship, if the robots work well, billions of robots in the commercial workforce.”

The revenue implication follows. If Figure captures even a small share of that wage base by replacing or augmenting human labor with humanoids, the resulting business is on a scale no current tech company has reached. “You’ll build like enormous, tens of trillions of revenue. You’ll build something massive,” Brett said. “Most tech companies trade at 10 or 20 times revenue. This is gonna be a huge business.”

Against that backdrop, the $39B valuation reads as a bet on category creation rather than a super duper stretched multiple. Software companies are priced against software TAMs. Figure is being priced against the global labor market, which is orders of magnitude larger than any vertical AI software opportunity. Brett’s view is that capital is not the constraint and the valuation is not the risk. The risk is execution on the underlying intelligence problem.

Brett wants Figure to be the first place artificial general intelligence is demonstrated in the physical world, not in a chat interface. “We think we have the recipe and we think we have the right training processes in place to do this. This year and next will be important to see if we can crack those.”

→ Listen on X, Spotify, YouTube, Apple

Subscribe now

BREAKING: The Anduril Thesis

Molly O’Shea — Wed, 29 Apr 2026 15:14:28 GMT

100 Years of Military History, Explained

, GP at Contrary, co-author of The Anduril Thesis, joins Sourcery to unpack the 300-page book he & Sachin Maini just spent two years writing on the most counter-positioned company in defense.

→ Listen on X, Spotify, YouTube, Apple

We go deep on a 100 years of military history — how Robert McNamara strangled America’s industrial golden goose in the 1960s, why cost-plus contracting incentivizes the worst possible outcomes, and how the U.S. ended up building $13 billion programs that turn into trillion-dollar overruns.

Then we map all of it onto Anduril: the Chick-fil-A founding story, why Lattice is the real product (not the drones), how the Sentry Tower beat a Lockheed contract that “defied the laws of physics,” the new $22B IVAS contract, undersea warfare and the cables that move $10T a day, and why ghost sharks are being built in Australia.

Kyle also breaks down Anduril’s funding history (including Contrary’s Series C entry), the talent density problem (only ~40 people in America can actually sell to the government), & the philosophy of conflict deterrence.

If you want to understand why Andúril is named after the Flame of the West, and why that matters for the next decade of geopolitics.. this is the episode.

📕 Get The Anduril Thesis by Kyle Harrison and the Contrary Research team: https://www.amazon.com/Anduril-Thesis-Kyle-Harrison/dp/B0GWLXQ7FC

Subscribe now

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

(00:00) Kyle Harrison, GP at Contrary & Co-Author of The Anduril Thesis
(01:22) The Anduril Thesis: A 300 Page Deep Dive
(04:52) How America lost its military edge
(07:33) The book that predicted the mess
(10:36) How bureaucracy broke the US military
(15:39) The ridiculous economics of modern war
(19:25) How Ukraine changed everything
(23:34) The Anduril thesis explained
(30:18) The pitch that changed defense forever
(36:52) Breaking down Anduril's tech stack
(45:26) The $10 trillion undersea threat
(57:43) Why founders need military mavericks
(01:03:53) Inside the new Department of War
(01:10:42) Anduril's unfair talent advantage
(01:21:20) When defense tech was toxic
(01:27:57) Why we must prepare for war
(01:39:27) The venture capital trap

Brought to you by:

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Merge—The leading provider of customer-facing integrations and agentic tools for frontier LLMs, Fortune 500 organizations, and B2B SaaS companies. Visit: https://merge.dev

The Anduril Thesis

Why America Forgot How to Build, & How One Company Is Rewriting the Playbook

A breakdown of our conversation with Kyle Harrison unpacking two years of research into one of the most counter-positioned companies in defense.

→ Listen on X, Spotify, YouTube, Apple

A 300-Page Book Born From a Hundred Years of Context

Kyle Harrison didn’t set out to write a book. Contrary Research, the public-facing arm of Contrary, where Kyle leads research, had been working through deep dives on companies like Stripe, OpenAI, and Databricks. Anduril was supposed to be next on the list. But something kept breaking the format.

“As we dug into it, we realized that in order to understand Anduril as a company, you genuinely have to understand like a hundred years of military history that it is being counter-positioned against.”

That realization turned a long-form research piece into a multi-year project involving a small army of research fellows, dozens of interviews, hundreds of white papers, and roughly 35 drafts. The team leaned heavily on two books in particular: The Kill Chain by Christian Brose (now Anduril’s President and Chief Strategy Officer), referenced more than 150 times, and Skunk Works by Ben Rich. Both lay the groundwork for what Kyle calls the most heartbreaking finding of the project, that America has been ignoring smoke alarms about its own industrial decline for decades.

“It’s just been multiple decades of… constantly tripping over ourselves into both the hubris of thinking that we have officially become the only great power left in the world.”

How America Strangled Its Own Golden Goose

At the end of WWII, the U.S. produced over 40% of global manufacturing capacity. The Department of Defense alone accounted for 36% of global R&D spending in the 1950s and 60s. Then came Robert McNamara.

McNamara, convinced the U.S. was overspending on defense, introduced PPBE, a planning and budgeting system that descended on places like Skunk Works as what Ben Rich described as a plague of auditors and bean counters. Spending dropped briefly. Then Korea and Vietnam happened, and the friction never went away.

“It’s not any one thing. It’s not any one tax. It is a million taxes of friction.”

The result: a defense apparatus that knows how to build big, exquisite, decade-late assets, and almost nothing else. The F-35 program, originally pegged at $13–15B, could ultimately cost up to a trillion dollars. And the cost overruns aren’t even the worst of it.

“It’s actually even worse than just a cost imbalance. You lack the iterative cycle to stay anywhere close to the cutting edge.”

China, meanwhile, has developed what are affectionately called “carrier killers,” $15–20M missiles capable of destroying multi-billion-dollar ships. War games modeling a U.S. vs. China conflict have, since roughly 2015, consistently ended with the U.S. losing. In a large-scale engagement, current estimates suggest the U.S. would run out of munitions in ~11 days, and rebuilding lost naval assets would take 25 years just to return to baseline.. yikes.

The Anduril Thesis: Counter-Positioning Everything

This is where Anduril enters. Kyle frames the company not as a typical tech startup, but as a deliberate inversion of every dysfunction in the existing defense system.

Where the DoD uses cost-plus contracting (which guarantees a percentage on top of whatever a project costs), Anduril uses fixed-cost. Where traditional primes ask the government to fund decades of R&D, Anduril self-funds. Where the Pentagon issues “built to spec” requirements, Anduril builds to mission.

“We absolutely want to listen to our customers because they understand their problem. We don’t necessarily listen to them about what the solution should be.” (Palmer Luckey, quoted by Kyle)

The incentives in cost-plus contracting are perverse by design:

“If the incentive is, no matter what it costs you, I will give you 10% on top of that, 10% of 3 billion is a much bigger number than 10% of 150 million. So as a traditional prime, you are deeply incentivized to reinvent the wheel from scratch every single time.”

The Sentry Tower is the cleanest illustration. A traditional prime had told the government their proposed border surveillance system would cost $5–10B, take 15 years, and probably defy the laws of physics. Anduril built one with off-the-shelf cameras, sensors, and batteries, and sold it first to a private rancher dealing with drug traffickers on his land before scaling to Homeland Security.

Lattice Is the Product (& Everything Else Is a Node)

People talk about Anduril’s drones, submarines, and munitions. Kyle argues they’re missing a very important piece.. Lattice, the underlying AI-driven connector.

“The real core product is… DARPA describes it as mosaic warfare. Chris Brose talks about it as a military internet of things. But basically it very much is this hub-and-spoke model of Lattice as an AI sensor-fusion-driven brain.”

Every Anduril product, Sentry Towers, Ghost drones, Altius loitering munitions, Roadrunner, the Dive line of autonomous undersea vehicles, Copperhead torpedoes, Fury autonomous fighter jets, the new Eagle Eye helmet from the $22B IVAS contract, exists to plug into Lattice. The strategy is to give every domain (air, land, sea, undersea) and every force level a node in the network.

Under-the-Sea

The undersea domain matters more than most people realize. Underwater data cables carry roughly $10 trillion in transactions a day. Many of them run through the Baltics, exposed. Anduril’s Seabed Sentry, Dive-XL, Barracuda, and Copperhead products are being purpose-built for a domain Kyle describes as practically harder to engineer for than space.

And the company keeps calling its shot. The recently announced Hyundai partnership for autonomous surface vehicles was teased at the end of the Copperhead launch video, which itself picked up exactly where the Fury video ended.

“From day one… it was about this software defined, hardware enabled web of products that can all work together.”

Founders, Mavericks, & the Mission Behind the Mission

Kyle keeps coming back to a pattern from defense history: every era of real military innovation pairs a founder-type with a “military maverick” willing to run interference against the bureaucracy. Bernard Schriver had Eisenhower. Skunk Works’ Kelly Johnson had Richard Bissell. William Moffett had Hoover.

What was missing for decades wasn’t the mavericks. There were plenty of admirals and strategists sounding the alarm in the 80s and 90s. What was missing was founder energy and institutional urgency. Anduril brought the first; rising tensions with China and the Russian invasion of Ukraine brought the second.

Co-Founders: Matt Grimm, Trae Stephens, Palmer Luckey, Brian Schimpf, Joseph Chen

But Kyle argues the most underappreciated part of the company isn’t its products or its politics. It’s its philosophy.

“If the mission is build cool stuff and blow stuff up, that’s not a mission. That’s a neat trick.”

He went back through every founder interview he could find. The word hope appears 44 times.

“Pacifism really is a privilege of the protected. There is somebody else who is willing to do those things and build those capabilities to protect our right to say, I don’t want to be involved in that.”

The company’s name is taken from Andúril, Flame of the West, the reforged sword in Lord of the Rings. Kyle quotes Faramir, the character Tolkien said he most identified with:

“I don’t love the blade for its brightness or sharpness. I don’t love the arrow for its swiftness or the warrior for his ferocity. I love only that which they defend.”

That’s the thesis underneath the thesis. Anduril’s only purpose isn’t to just win wars. It’s trying to make them defensible & advanced enough not to happen in the first place, & to turn allies like Taiwan into what Palmer Luckey calls “prickly porcupines.”

The Funding Story & Vibe Shift

Anduril’s funding history maps almost exactly onto America’s slow awakening. In 2017, working in defense was deeply unpopular. Engineers ghosted Palmer Luckey when he revealed what he was building, and Elad Gil reportedly received angry messages from friends for leading an early round. Founders Fund and General Catalyst were the early believers. Andreessen came later. Contrary entered at the Series C and has invested in every round since.

Then came what Palmer calls his “I told you so tour.” 2022, the Russian invasion, the post-ZIRP collapse, the public market reset, was a one-two punch followed by a real takeoff. Anduril had been saying since 2017 what Putin had said publicly that same year: whoever controls AI will control the world.

Was the acronym is hiding in plain sight? AI… Anduril Industries.

Hot Take

But Kyle warns that the vibes will swing again, and that’s not just an Anduril problem. It’s a venture problem. Which sets up his closing hot take, delivered against his own self-interest:

“My hottest take is that fewer people should be raising venture capital… There is a huge swath of companies that could be very good companies that are very bad venture-backed businesses.”

He points to a friend running a metal-shelving manufacturer doing $17M in revenue and $3M in profit, growing 3% a year, living a great life. A terrible venture outcome. A wonderful business. The deeper point is that the next phase of American reindustrialization, supply chains, manufacturing, dual-use industrial capacity, won’t all fit through the venture funnel. And it shouldn’t have to.

“Every single time, military superiority is front-run by industrial superiority. We did that in World War II. The problem is now that China is doing that today.”

The book is The Anduril Thesis, by Kyle Harrison and the Contrary Research team. If you want to understand the next decade of geopolitics, it’s a useful place to start.

→ Listen on X, Spotify, YouTube, Apple

MORE ANDURIL …

Subscribe now

BREAKING: Skydio Commits $3.5B to U.S. Drone Manufacturing

Molly O’Shea — Fri, 24 Apr 2026 15:07:57 GMT

Welcome to Dronesville, USA

On top of Skydio’s $110M Series F at a $4.4B valuation announced yesterday, Adam Bry, CEO of Skydio, announces a $3.5 billion commitment to expand American manufacturing over the next 5 years.

To celebrate, we spent the day at Skydio’s San Mateo headquarters for a full tour of their facilities to see what they’re about to 5x.

→ Listen on X, Spotify, YouTube, Apple

The $3.5B announcement involves a new US facility five times the size of the current one (Skydio’s fifth expansion in eight years), over 2,000 new Skydio jobs, 3,000+ additional roles across the US supply chain, and more than $1 billion directed to domestic suppliers under a new program called SkyForge. Skydio already manufactures more dual-use drones than any company outside of China. This is the bet to keep it that way.

The backdrop: 60,000 drones shipped, 3,800+ customers, 1,200+ public safety agencies, every branch of the US military, & 29 allied nations. Skydio DFR now arrives on scene first 71% of the time & resolves nearly 1/4 of calls without dispatching a patrol unit.

“U.S. innovation invented the airplane, ramped up manufacturing to win WWII, put a man on the moon, broke the sound barrier, & commercialized space travel. Skydio has proven that American companies can compete & win in the civilian drone market against products from our adversaries.” — Adam Bry

On the tour, we visit:

Rooftop dock array (drones as cloud infrastructure)
DFR Command, flown live
The “wind wall” stress-testing hardware 24/7
R10, the new indoor tactical drone
F10, a fixed-wing thrown and caught by a robotic arm
X10 water rescue demo with the auto-inflating rest tube

Subscribe now

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

(00:00) A flight every 30 seconds
(00:57) Rooftop docks
(03:26) Tour roadmap
(04:42) Adam's RC airplane origin story
(06:10) Flying a dock drone, live
(10:04) The "force multiplier" moment
(12:16) Multi-drone operations
(14:17) Unarmed sensor platforms and the China context
(15:29) Why community acceptance surprised Adam
(19:31) The wind wall

(20:40) R10, built for indoor tactical
(23:30) Flying R10 through a barricaded-suspect scenario
(27:30) Reading oil and gas gauges with a drone
(29:26) F10, the fixed-wing caught by a robotic arm
(30:40) X10 water rescue demo

Brought to you by:

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Merge—The leading provider of customer-facing integrations and agentic tools for frontier LLMs, Fortune 500 organizations, and B2B SaaS companies. Visit: https://merge.dev

America’s Largest Drone Manufacturer

The $3.5 Billion Commitment

On April 24, 2026, Skydio announced a $3.5 billion domestic investment over the next five years, spanning manufacturing expansion, R&D, and US supply chain development. The commitment funds over 2,000 new Skydio jobs, supports 3,000+ additional roles in the supply chain, and directs more than $1 billion to domestic suppliers through a new program called SkyForge. The accompanying facility expansion will be 5x the size of Skydio’s current footprint, representing the company’s fifth manufacturing expansion in eight years.

The strategic message is explicit. Skydio already manufactures more dual-use drones than any company outside of China, and this commitment is designed to widen that gap. The SkyForge model is structurally interesting: select suppliers will be invited to co-locate production capacity with Skydio, gaining access to Skydio’s engineering talent. That is not a standard supplier relationship. It is vertical integration expressed through partnership, and it solves the component-level supply chain fragility that has plagued US hardware manufacturers competing against Chinese vertical stacks.

“U.S. innovation invented the airplane, ramped up manufacturing to win WWII, put a man on the moon, broke the sound barrier, and commercialized space travel,” Adam Bry said in the announcement. “Skydio has proven that American companies can compete and win in the civilian drone market against products from our adversaries.”

The Current Scale

The company operates at a scale that most market observers still underestimate. Skydio has shipped more than 60,000 drones, logged over 3.4 million customer flights, and currently runs 1.5 flights per minute in non-defense markets alone, representing 2,000+ flights per day, up 278% year-over-year. Approximately 900 employees support an 80,000 square foot manufacturing facility in Hayward, California, with headquarters in San Mateo. Annual revenue is in the nine figures, with rapid growth.

The customer base is the more defensible asset. Every branch of the US military. Twenty-nine allied militaries. 1,200+ public safety agencies across 42 states, a figure that doubled year-over-year. 1,200+ critical infrastructure companies including 250 utilities (Dominion, PG&E, Duke, Tennessee Valley Authority, National Grid, Berkshire Hathaway, among others). 450+ energy companies. 45 of 51 US state transportation agencies. 600+ state transportation, civil engineering, and construction organizations. 11 million Americans now live within two miles of a Skydio DFR dock.

Adam’s framing captures the category dynamic concisely: “People don’t really understand how deployed and impactful these things already are.” The industries Skydio serves do not produce viral consumer moments. They produce compounding enterprise revenue in the background of critical infrastructure, which is the profile of a business that gets underestimated until market share is locked in.

The Tour: Hardware as Operating Discipline

The San Mateo campus is organized around a set of physical demonstrations that double as operational capabilities. The rooftop hosts an active dock array in two configurations: an endurance rig that flies unattended 24/7 as a reliability test, and a customer demo rig that global sales teams log into remotely to run representative missions. Adjacent to the rooftops sits the wind wall, an array of propellers generating 30 to 35 mph sustained winds against docks in multiple orientations, with variants built to add rain.

The R10 development building hosts a simulated barricaded-suspect house used for live indoor tactical flights. Outdoors, a purpose-built oil & gas inspection rig, including substation equipment, lets autonomous drones practice reading steam gauges and detecting leaks. The F10 fixed-wing prototype is tested with a robotic arm launch-&-catch system in a dedicated bay. The X10 water rescue demo uses an auto-inflating rest tube dropper, built in response to NYPD requests for coastal rescue capability.

“The number one feature for us is reliability, making this stuff work end to end, bulletproof day, night, wind, rain, that’s the hard part… You can sort of think of this as like continuous integration for our development process with real autonomous hardware in the loop.”

The Product Line

Skydio now sells specialized airframes for specialized missions, unified by a common autonomy and software layer. The X10 is the general-purpose workhorse compatible with the autonomous dock system, carrying a three-axis gimbaled sensor package capable of reading a license plate at 800 feet and detecting a person at a couple of miles. Attachments include spotlights, speakers, and the rest-tube dropper. The X10D is the militarized variant, fielded under the US Army’s Short Range Reconnaissance Tranche 2 program, GPS-denied capable, integrated into ATAK mission command.

The R10 is the new indoor tactical drone, designed specifically for confined and dangerous spaces. Headlights for dark interiors, speaker and microphone for two-way communication, form factor small enough to fly through doorways. Adam framed the use case during a live flight: “you could have an armed suspect in there and rather than sending a cop in who’s either gonna get shot at and maybe have to shoot back, you send the drone in.” R10 is already in customer hands and enters general availability in the coming weeks.

The F10 is the fixed-wing prototype built for long-range, high-speed missions such as utility line inspection and large-jurisdiction public safety. “If you want to go fast, if you wanna go far, you need a wing. It’s just a much more efficient way to generate lift.” The robotic arm launch-and-catch mechanism is how F10 stays inside the dock-and-remote-operation paradigm that makes the rest of the fleet economically viable.

The platform layer underneath is where the unit economics diverge from legacy drone businesses. DFR Command is the software most public safety customers experience the product through: map-based incident dispatch, one-click drone launch, autonomous wayfinding via Skydio’s Pathfinder system, multi-drone operations from a single interface. The Transparency Portal lets agencies publish their flight logs to the public. Docks provide thermal regulation, network connectivity, and charging, turning drones into, as Adam put it, “a little bit just like using a cloud server.”

The horizontal nature of the stack matters: “It’s pretty horizontal technology. It’s like building a computer or something. You can use it for all different kinds of things. It’s really the software that defines how you interact with it and what the capabilities are.”

DFR: The Application Driving Adoption

Drone as First Responder is the fastest-moving deployment in the portfolio and the one producing the clearest outcome data. A Skydio internal analysis of 6,000 flights across 61 public safety agencies documented the drone arriving first on scene 71% of the time, 23% of calls cleared without dispatching a patrol unit, and helpful intelligence provided on 92% of flights. SFPD credits the program with a 42% reduction in auto theft, 500+ arrests, and a 30% drop in overall crime.

Lee County, Florida deputies located a missing elderly man in under ten minutes. Bernalillo County deputies de-escalated a call involving two armed children (ages 7 and 9) without resorting to deadly force. OKCPD and OKCFD located an injured man near a rail line in time to save his life.

Adam’s framing of the rate of deployment captures the inflection: “on average now in public safety, our customers are flying once every 30 seconds… this is across the nation 24/7.” That pace reflects a product category moving from pilot programs to default tooling. The agency count doubled year-over-year, from 600 to 1,200+, across 42 of 50 states. The addressable base continues to expand, and adoption compounds as neighboring agencies see results.

The non-obvious insight, and the one that matters most for market sizing, is that community acceptance has been the gating factor rather than technology. “Probably the single biggest, I would say, positive surprise that I’ve had over the last five years is the level of community acceptance and buy-in and really excitement for the law enforcement application of these things.”

Adam’s argument is that transparency is the unlock: “the best answer to privacy is actually transparency as well.” That policy wedge, expressed through the Transparency Portal and proactive agency communication, is not easily replicated by foreign competitors and represents a durable advantage in US public safety.

Site Security

The fastest-growing application in the Skydio portfolio is not public safety. It is site security, driven primarily by data centers and critical infrastructure operators. The reason matters for understanding the unit economics. A single docked drone at a data center can run incident response, perimeter patrol, and infrastructure inspection (power, cooling) from the same asset on the same day. That is one capital expenditure amortized across multiple revenue-generating jobs, which is structurally different from single-purpose hardware.

The same logic extends across the customer base. Utilities already using Skydio for transmission line inspection can add substation security as a software decision. Construction firms doing site documentation can add perimeter security without adding fleet. Every additional mission loaded onto an installed dock increases customer value without incremental hardware cost.

“These are unarmed drones. These are basically flying sensor platforms. And we sell very similar products to the military. Our class of drone to the military is what they call ISR, Intelligence Surveillance Reconnaissance.”

A flying sensor platform that can be redirected via software to new use cases is a different asset class than a drone that does one thing. Investors evaluating the company should be pricing the platform, not the individual airframes.

The National Security Thesis

The national security argument is the one that connects the $3.5 billion commitment to the broader strategic context. The US is investing billions in one-way attack drones, but as Skydio’s positioning states: “scale alone doesn’t deliver dominance. The missing piece is persistent, organic ISR at the tactical edge. Without a survivable ‘hunter’ platform maintaining target custody, adding more ‘killer’ effectors produces more sorties but fewer confirmed effects.” The X10D is positioned as that hunter: GPS-denied capable, autonomy-driven, integrated into ATAK.. “you can’t dominate what you can’t see.”

Commercial validation has arrived. The $52 million-plus US Army order for X10D under SRR Tranche 2 is the largest single-vendor tactical sUAS order in Army history. Combined with deployment across every US military branch and 29 allied nations, this establishes Skydio as the default Western alternative in a category where the default has been Chinese for a decade. The dual-use posture matters: the same autonomy stack, reliability discipline, and manufacturing scale that serve public safety and utilities are what the military is buying, which is why the R&D investment compounds across both markets.

Dronesville, USA

The closing frame from the tour applies as well to the investment thesis as to the product roadmap. “This campus is slowly evolving into Drone Mecca. Our goal is to just have this be the most exciting place in the world.” The $3.5 billion commitment, $110M investment at $4.4 billion valuation, Army contract, doubling DFR adoption, and the specialized product line are all expressions of the same underlying bet: that autonomous flying robots will become critical infrastructure for the industries that run America, and that the company furthest along the reliability curve wins the category. Skydio is making that bet visible, funded, and domestic.

→ Listen on X, Spotify, YouTube, Apple

Subscribe now

BREAKING: Skydio Hits $4.4B on $110M Series F

Thu, 23 Apr 2026 15:04:26 GMT

Why Skydio Raised Less Than Investors Offered

Today, Skydio announces a $110M Series F at a $4.4 billion post-money valuation. CEO Adam Bry joins Sourcery to dive into the oversubscribed funding process, scaling to over 9 figures in revenue (!!), commercial progress, military + public safety use cases, & their manufacturing edge at Skydio's Hayward factory.

→ Listen on X, Spotify, YouTube, Apple

ICYMI: Skydio is the largest drone factory in the US at scale production.

“The most significant fact in our Series F is how little we are raising. Despite investor demand to put substantially more into the company… our capital needs are rapidly decreasing.” – CEO Adam Bry

Skydio by the numbers:

$100Ms in annual revenue, strong unit economics
~900 employees
60,000 autonomous flying robots shipped
3.4M+ customer flights
3,800+ enterprise customers
1,200+ public safety agencies (doubled YoY)
45 of 51 state transportation agencies
Every branch of the US military plus 29 allies
11M+ Americans live within 2 miles of a Skydio DFR Dock

In this conversation, we cover:

$100M Series F & the unit economics behind a $4.4B valuation
$50M, 3,000 drone US Army contract and tripling production in 2026
R10 indoor drone & F10 fixed wing
Turning down DJI’s acquisition offer in 2014
Getting sanctioned by the Chinese government in 2024
Drone as First Responder: Public Safety customers in 42/50 states, 42% auto theft reduction in SF
Built a China independent supply chain & US manufacturing at scale
Flying agentic AI & the future of autonomous robotics
Adam’s hottest take on the robotics hype cycle

If you’re interested in reindustrializing America, drones, AI, robotics, & fundraising in hard tech, this one’s for you.

Subscribe now

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

(00:00) Adam Bry, CEO of Skydio
(00:54) Inside Skydio Factory
(01:31) Series F & Scaling
(03:17) Customers And Growth
(04:10) Origins & Reliability
(08:04) Civilian Missions Focus
(09:19) Drones For Public Safety
(12:29) Edge Compute & Partners
(14:20) Expanding Drone Lineup
(17:10) Beating China & DJI
(19:34) US Supply Chain Resilience
(21:33) Reindustrialization Timeline
(22:35) Hiring & Leadership
(27:12) AI Agentic Drones
(30:08) Metrics & Operating Model
(31:49) Hottest Take And Advice

Brought to you by:

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Merge—The leading provider of customer-facing integrations and agentic tools for frontier LLMs, Fortune 500 organizations, and B2B SaaS companies. Visit: https://merge.dev

The Flying Robot Company That Raised Less Than Investors Wanted

Conversation with Skydio CEO Adam Bry on Series F, Autonomy, & the New Physics of Robotics Capital

Most venture headlines in AI and robotics run the same way: more capital raised, more dilution accepted, more runway needed to chase receding horizons. Skydio’s announcement this week runs the other direction. $110 million at a $4.4 billion post-money valuation, a modest check by current standards, deliberately so.

Turns out, when you have strong unit economics + 9 figures in revenue you can think more critically about preserving equity & not taking money just for money’s sake.

→ Listen on X, Spotify, YouTube, Apple

Twelve years in, Skydio has moved from the first drone hype cycle, where “a bunch of capital was invested,” to being, in Bry’s telling, “really the only surviving company that’s building drone hardware from that era.” What makes the round interesting is how Bry himself frames it: “I really think the most significant fact in this whole thing is how small it is. We’re in the very rare and hard earned position amongst robotics and AI companies of actually having rapidly declining capital needs.”

It’s a philosophy he traces back to something older than the current cycle. “One of the frameworks I think about is get rich slowly,” Bry said. “Enduring value and enduring impact is built on a lot of patience and a lot of hard work... the signal trumps the noise. It always does.” What follows is our read on the business, the product thesis, and why the scarcity of this raise may be the loudest signal in the release.

A Humble Series F

The core business is funding the core business.. Skydio reports nine figures of annual revenue, 60,000 flying robots shipped, and roughly 4,000 enterprise customers, including every branch of the U.S. military, 1,200+ public safety agencies, and 45 of 51 state transportation agencies. In a sector pricing in optionality rather than operating income, that reads differently.

The capital is going to unglamorous work. “The number one constraint we’re facing right now is building more drones faster,” Bry said. “We will be tripling production over the course of this year.” A $52 million, 3,000-drone Army order, “actually not that much of an outlier for us these days,” tells you throughput is the bottleneck. Raising only what you need when the market would give you more is a rare signal in robotics.

Autonomy Bet, Compounding 12 Years Later

Skydio’s contrarian call was that drones needed to fly themselves. That bet nearly ended the company: DJI expressed acquisition interest in 2014. Bry declined because “we felt like they didn’t really understand the importance of autonomy and how much that was gonna change, who could use drones and how they could use them.”

The harder insight was about reliability, not capability. “It’s very easy to make a demo, like getting a drone to take off and fly around manually is like 0.1% of the way there,” Bry said. “What’s really hard is... making it work 90% of the time, 99% of the time, 99.9% of the time, like chasing those nines of reliability.”

The analogy he reaches for is self-driving: “Google could demo a self-driving Waymo like 15 years ago, but it’s taken them 15 years to make that reliable.” Skydio’s nines, already “built on hundreds of thousands, millions of flights,” now power what Bry calls a “family of flying robots,” including an indoor drone and a fixed-wing aircraft. Twelve years of narrow focus produced a platform.

Public Safety: Business That Also Saves Lives

Skydio drones now sit in 1,200+ public safety agencies in 42 states, a customer count that doubled year over year. Internal analysis of 6,000 flights across 61 agencies found the drone was first on scene 70% of the time, 23% of calls were cleared without patrol, and the drone provided helpful intelligence 92% of the time.

Bry’s prediction: “I think a default expectation in five years is if there’s an emergency, you call 9-1-1, a drone shows up in a few seconds and that’s gonna be everywhere in the US, hopefully everywhere in the world. And that’s gonna change the way policing works.”

He frames these as transparency tools, not surveillance: “these things are essentially flying body cameras.” The evidence spans the stakes spectrum.

San Francisco has seen a 42% reduction in auto theft since launching its drone program. A rooftop call in SF reported a rifle; the drone arrived first and saw a broomstick. Less than 1% penetration against total 9-1-1 volume gives you a sense of the runway.

The China Problem, & Why the Supply Chain Became a Feature

“We have the great joy and we’ve had it for the last 12 years of being the US underdog against the best of Chinese industry, really state sponsored Chinese industry,” Bry said. Rather than trying to out-spec DJI on hardware, Skydio chose autonomy and end-to-end solutions as the competitive axis.

Then the geopolitics turned operational. “A year and a half ago, we had the great honor of being sanctioned by the Chinese government,” Bry said. “They announced the sanctions and then they showed up at the suppliers that we still had in China, shut them down, stopped them from doing business with us, really tried to, to kill us.” The result: “We now have by far, the most secure drone supply chain in the world, independent from China, that’s operating at increasing levels of scale.” In a market where critical-industry customers are writing Chinese drones out of procurement, being the non-Chinese manufacturer at scale isn’t a marketing line. It’s a category position.

Building “Shenzhen” in the East Bay

Reindustrialization is slower than the rhetoric suggests. “It’s not as fast as people might like,” Bry said. “But it’s not an impossibility. It really just is a matter of like getting the flywheel spinning.” His mental model is ecosystem density. “There’s some alternate universe where this looks like what Shenzhen looks like, and I don’t think there’s anything fundamentally that should stop us from getting there.”

Skydio is positioning as an anchor tenant: an 80,000 square foot facility in Hayward, roughly 900 employees, and a US manufacturing bet placed in 2014, long before “reshoring” became a policy talking point.

The talent math is where it gets interesting. “The talent level’s great at every skill set you need,” Bry said. “There’s not as much of it as we might like, but the talent level is incredibly high.” A meaningful share comes from one familiar source:

“Tesla has generated some great alumni & we have a lot of fantastic folks that have come from forged in the fire of production hell on Model 3.” Seen throughout our interviews Tesla has served, time and again, as the involuntary training program for a generation of American hardware operators.

The hiring filter is unconventional by design. “We just focus on people that we think have enormous potential, whether or not they have sort of like the proven, exact right experience.”

What Bry screens for: “that sort of spark of capability... the combination of the fire and passion to do hard things and then the horsepower to get it done.” The closing pitch is structural: “There’s no other company in the world really, that has the combination of like true cutting edge sci-fi technology... in a real business that’s really working with customers that depend on this in like lifesaving critical missions.”

Flying Agentic AI, & Why Hardware Makes It Harder

The product roadmap, in Bry’s framing, points toward “our drones being like flying agentic AI, just like you have an agent that you interact with on your computer or in the cloud. This thing is an agent that can move & do work for you in the physical world.”

What makes this more than narrative is the underlying infrastructure, “a very rich simulation environment” that lets the team “test and iterate with the real physics that you care about in the loop. So you can do a lot of like the speed of software with hardware involved.” That last phrase is the thesis in one line.

Bry is unsparing about where the sector is heading. Asked for his hottest take:

“I think that we are gonna see a lot of pain and carnage on the investments and expectations that are being set in this sort of like most recent round of robotics companies. ‘Cause it’s just, the physical world does not work the way the software world works... The learning cycles are slower, they’re more expensive.”

A CEO who has already survived one drone trough of disillusionment naming the next one out loud is worth noting.

How Skydio Operates

Bry runs against cascading OKR culture. “I am not big on these like sort of like super granular cascading OKRs... organizing really small teams around them. You want the team to be as small as it possibly can be. And in the AI you can do pretty amazing things with pretty small teams.” The weekly executive meeting opens with customer success, “good, bad, and ugly.”

His advice to founders building into the same bubbly moment is pointed: “get really anchored in the customer problem that you’re trying to solve and building world class technology to solve it.” Easy to say. Much harder when the gravitational pull of fundraising milestones and vanity metrics is core to silicon valley culture.

What to Watch

Skydio’s Series F is a rare artifact in the current cycle: a priced round sized to operating reality, not market appetite. The thesis Bry laid out, autonomy as the compounding moat, public safety as the growth vector, US manufacturing as a structural advantage, and flying agentic AI as the frontier, is as coherent as any in robotics this year.

The risks are the ones Bry named himself: the physical world moves slower than software, reliability is earned one nine at a time, and a lot of peers in this cycle will learn that the hard way.

Near-term milestones are concrete.

Triple production. Scale DFR 3x. Ship the indoor drone and fixed-wing. Extend the end-to-end playbook from law enforcement into site security, utility inspection, and battlefield ISR. If Bry is right that “the signal trumps the noise,” this raise and how it was constructed is one of the clearer signals the sector has produced in a while.

→ Listen on X, Spotify, YouTube, Apple

Subscribe now

BREAKING: General Catalyst's First-Ever Quarterly Review

Mon, 20 Apr 2026 15:19:34 GMT

Is GC Going Public??

Hemant Taneja, CEO of General Catalyst, joins Sourcery to unpack GC’s first-ever quarterly review.

→ Listen on X, Spotify, YouTube, Apple

We cover why he wrote it, the state of GC’s $43B ecosystem, why the software buyout model is broken, Anthropic’s Series G at $380B, the $5B India commitment, buying Summa Health in Akron, the Anthropic–DoW controversy, Mythos & Project Glasswing, Nikesh Arora joining the GC board, and his take on “Tech Lash 2.0.”

Plus: Is GC going public??

HUGE thank you to Deel Co-Founder & CRO Shuo Wang for the introduction to Reggie & the GC team, I am SO happy we were able to make this happen so quickly! What a treat.

Read the Q1 2026 Review (overview below)

Subscribe now

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

(00:00) Hemant Taneja, CEO at General Catalyst
(01:21) Present state of General Catalyst
(05:08) Seed, Creation & the Customer Value Fund
(06:54) The purpose behind GC’s quarterly review
(12:26) The biggest AI themes right now
(15:06) How AI is changing investing decisions
(19:25) Why seed and Series G require the same conviction
(22:55) Why does he holds two titles at once?
(25:17) Why Silicon Valley glorifies the wrong founders
(27:52) Building an AI-native hospital
(30:20) Managing funds through bubbles and hype cycles
(33:13) How funds sell ownership stakes
(34:59) Why the software buyout model is broken
(38:44) Vista's new software buyout debt fund
(39:20) Global Resilience vs. American Dynamism
(41:49) GC’s $5B bet on India
(45:48) What global investing teaches you
(47:09) The four forces of tech lash
(50:12) Anthropic and the Pentagon
(54:06) Is GC going public?
(55:26) The Puka shell necklace era

Brought to you by:

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Merge—The leading provider of customer-facing integrations and agentic tools for frontier LLMs, Fortune 500 organizations, and B2B SaaS companies. Visit: https://merge.dev

GC’s First-Ever Quarterly Review

CEO Hemant Taneja on AI’s Dominance, Broken Buyouts, & the Gray Areas of Silicon Valley

I just sat down with Hemant Taneja, CEO of General Catalyst, at the top of GC’s San Francisco tower to walk through the firm’s first-ever quarterly review.

It’s a deliberate break from the X-thread, fireside-soundbite, click-bait rhythm that dominates venture commentary, a slower, more honest attempt to make sense of a quarter where, as Hemant puts it, “a lot of large things happened simultaneously, none of them finished.”

Below, I’ve pulled out the most data-oriented sections of the review and our conversation, the software buyout math, the Anthropic trajectory, the India commitment, the CVF credit rating milestone, and tied them back to the broader argument Hemant is making about where durable value lives in the AI era.

→ Listen on X, Spotify, YouTube, Apple

Why the First Quarterly Review, & Why Now?

Hemant’s explanation for the review’s existence is itself a thesis about the industry. The formats venture has defaulted to, X, LinkedIn, conference stages, reward certainty & controversy. None of them, in his view, help anyone actually make sense of the world.

When I asked him why he started writing it now, he told me, “A lot of that generally skews us towards taking very polarized views, because the stuff that amplifies in the world tends to be very black or white, & the progress gets made in the gray.”

The review itself opens on the same note, the volume of happenings, the heat of the discourse, and the open question mark of what the very near future will look like can leave one feeling confused. Hemant’s answer is to slow down, once a quarter, take stock, and be willing to sit with complexity rather than reach for a clean narrative.

The format also forces GC to be honest about the arc it's actually on. In our conversation, Hemant described three core investment strategies:

Seed (the firm's venture capital fund, which he calls the largest seed fund of its kind)
Creation (the build-and-transform strategy run by Marc Bhargava that includes AI-native rollups like Long Lake, Crescendo, Udia, and company hatches like Hippocratic AI)
Customer Value Fund (the investment-grade-rated product for scaling proven businesses)

But those three sit inside a larger ecosystem. Percepta is GC's in-house AI engineering and research company, bringing applied AI into GC's portfolio and partner institutions. HATCo is the healthcare transformation arm behind the acquisition of Summa Health in Akron, Ohio, where the firm is trying to build the first AI-native community hospital. GC Institute is the policy arm operating out of DC, Brussels, and Delhi. And GC Wealth, launched publicly in early 2025 under former First Republic executive Dave Breslin, is the firm's full-service wealth management business for founders and entrepreneurs in GC's Famiglia.

The Anthropic Trajectory & What “Beneficent Singularity” Actually Means

The review leads with AI because, as Hemant writes, “I can’t start anywhere else.” The central data point: Anthropic went from $1B to over $20B (now allegedly at $30B ARR) in annualized revenue run rate in fifteen months. GC first invested in March 2025 at a $61.5B valuation. The March 2026 Series G valued the company at $380B, one of the largest venture rounds in history.

Hemant frames this through the Collisons’ 2026 Stripe letter, which reported over 700 AI agent companies launching on the platform in a single year. Their conclusion: the world may be at the advent of a different and “hopefully much more beneficent singularity.” Hemant lingers on those word choices, hopefully, beneficent, because they frame the singularity as an aspiration, not a foregone conclusion.

His point is that capability is no longer the binding constraint. Trust is. The companies that will matter in thirty years, he argues in the review, are not necessarily the most technically capable ones. The actual transformation happens inside regulated, high-stakes environments, healthcare, education, defense, financial systems.. he makes it clear, those environments don’t reward or value raw capability. They reward embedded trust.

That reframing is what drives GC’s full-arc positioning: 20 seed investments this quarter alongside the Anthropic Series G, Percepta building AI engineering capacity in-house, and Creation doing AI-native rollups of incumbent businesses.

Why the Software Buyout Model Is Broken

The most data-dense section of the review is Hemant’s breakdown of software buyout economics, and it’s worth walking through carefully. The setup: a high-quality B2B software business bought a few years ago with sticky enterprise contracts, 90% gross margins, and 25% annual growth. Entry price: 25× EBITDA (roughly 40–50× FCF), partially financed through private credit, underwritten to a five-year hold. Over that hold, EBITDA grows 50%. The debt comes down. By every operational measure, the deal worked.

And yet the equity doesn’t. EBITDA growth added $625M in enterprise value. Multiple compression from 25× to 12× destroyed $975M. The deal returned 0.68× MOIC, roughly a third of LP capital lost on a deal where the underlying business succeeded. Annualized over 5 years, that’s a deeply negative IRR of approximately -7%.

When I asked Hemant to walk me through this live, he put it simply: the assumption of terminal value is gone.

“In the world of AI today where code is self-writing, to say that a piece of software in a company that has been existing, let’s say five years, the free cash flow, that is worth 30 times, means you’re gonna have 30 years of generating that free cash flow is what you’re paying for. How can you ever make that assumption when technology is changing so fast to say these things are that durable?”

His path forward: structure software buyouts the way mature-category buyouts have always been structured, on cash flows, not exit multiples.

In the review, he runs the counterfactual. Same $50M EBITDA business, 25% growth, but entered at 10× instead of 25×. $500M enterprise value, 3× EBITDA leverage, $350M equity check. Over five years, EBITDA grows to $75M through genuine AI-driven operational improvement. Exit clears at 14×. Enterprise value: $1.05B. Equity recovered: $950M. That’s a 2.7× MOIC and an IRR of approximately 22%.

The CVF Credit Rating & What It Unlocks

The other major financial milestone Hemant discloses in the review is that GC’s Customer Value Fund reached an investment-grade credit rating this quarter, BBB or higher on roughly two-thirds of the capital. To the best of GC’s knowledge, no venture product has ever reached that threshold before.

The unlock is scale. Apollo Global Management runs close to a trillion dollars in AUM, roughly 80% of it in investment-grade credit strategies. That’s nearly $800 billion in capital at a single asset manager that venture has historically been unable to access, because venture investments carry no defined duration or structure. CVF, which launched in 2021, can now bring that kind of capital to bear for GC’s companies, and it did so in under five years.

Fast-growing companies like Fivetran, Kandji, Ro, Travelperk, Superplay, & Upside are already using CVF to fund customer acquisition costs, with CFOs describing it as a dedicated growth balance sheet that scales with unit economics rather than market conditions.

In our conversation, Hemant tied this back to the software buyout argument. CVF gives GC the ability to drive returns in software on a cash flow basis, without the terminal-value dependency that’s impaired so much of the current vintage. Paired with Creation (for transforming established assets) and Percepta (for the AI orchestration layer), it’s the financial infrastructure behind GC’s thesis that the next era of private markets returns comes from AI-driven top-line growth, not cost-cut exits.

GC’s direct exposure to pure SaaS, for context, is approximately $4 billion out of $43 billion in AUM, much of which sits across strong pre-2022 companies in fintech, healthcare, and defense that have embraced AI and are scaling faster than ever.

The $5B India Commitment & Global Resilience

In February, Hemant flew to New Delhi for the India AI Impact Summit and announced a $5B GC commitment to India over the next five years, one of the largest single-country investment commitments in venture history. When I asked him why India, he broke it into three forces playing out at once.

The first is systems. India has a pattern of solving for constraint: generic pharmaceuticals, UPI (which now processes more transactions than Visa and Mastercard combined), and Aadhaar (the largest biometric database on earth).
The second is scale, India will be the third-largest economy within the decade, with one million young people entering the workforce every month.
The third is resilience: India is actively diversifying away from Russian defense procurement, with 75% of modernization spend now going domestic (up from roughly 60% five years ago) and defense exports up 34x over 11 years.

The India commitment sits inside a broader global resilience thesis. Hemant told me in the interview that the geopolitics will force supply chains to shift, Europe won’t buy American defense products at scale, India won’t rely on Russia and America, so defense primes will emerge everywhere.

GC is already invested in Anduril & Saronic in the U.S., Helsing in Europe (Jeannette zu Fürstenberg, Daniel Ek, & Paul Kwan), and Raphe in India.

In Hemant’s framing, this isn’t just TAM expansion. It’s a values statement about inclusive prosperity: if AI concentrates opportunity in a handful of companies in a handful of countries, society will reject it, and the whole transformation stalls.

Anthropic–DoW Controversy & Mythos

The part of the review I most wanted to push him on was the Anthropic–Department of War situation from February. The public discourse simplified fast into “morally superior AI vs. unbridled acceleration.” Hemant’s position is that neither side captures what actually happened.

“I cannot say the perspectives on either side were wrong,” he told me. “It’s just like from where you sit and how you think about the world and what’s important is really what led to those issues. But what it shows you is just the complexity of what we have to go through.”

His concern in the review is less about the specifics of that deal and more about what it revealed.. how quickly the public reached for the exit in the face of genuine complexity. The same questions, where should AI have no role, who decides, will show up in healthcare, education, workforce, and online safety, and the industry needs governance architecture, not leadership preference.

Mythos is the other side of that coin. Anthropic could have released it broadly; instead they rolled it out through Project Glasswing to a tight group focused on eliminating vulnerability debt in critical infrastructure. Hemant calls the restraint admirable, “the opposite of move fast and break things,” and adds news that Nikesh Arora, who’s joining GC’s board, is part of the “Glasswing 40.” His broader point is that the individual moral frameworks from leaders like Dario aren’t a governance system. The industry needs the architecture to make responsible deployment practice rather than preference.

The thread running through all of this, Anthropic, Mythos, Iran, the DoW, is Hemant’s argument that the world is irreducibly gray, and the leaders worth following are the ones who can stay in the gray long enough to design something durable.

Nikesh Arora Joins the GC Board

One of the more consequential governance moves Hemant discloses in the review is that Nikesh Arora, CEO of Palo Alto Networks and one of the “Glasswing 40” working with Anthropic on Mythos, is joining GC’s Board. In the review, Hemant writes:

“I’m very glad that Nikesh Arora, who is joining GC’s Board, is part of the ‘Glasswing 40’. Nikesh has led some of the most consequential technology organizations in the world, and I’ve long admired the clarity and ambition he brings to everything he does. His presence reflects the caliber of judgment being brought to bear on a model that could either do serious damage to or rigorously armor the world’s critical systems.”

Ken Chenault remains Chairman, and Nikesh is joining as lead independent director. When I asked Hemant about the appointment in our interview, he laid out the reasoning:

“He’s been, first of all, incredibly brilliant. He’s an entrepreneurial soul running a Fortune 500 company, & he’s done an incredible job building that. He’s also a great investor mind. So as I think about our GC ecosystem, which is a collection of investors & builders, I just think he’d be a great mentor for all of us.”

The framing mattered. Hemant isn’t treating this as a prestige appointment. He’s treating it as another node in the leadership structure he’s built around himself, Ken Chenault as Chairman, former industry CEOs like Ken Frazier, Kate Walsh, Steve Klasko, and Daryl Tol helping run the company, and now Nikesh as a sounding board on where AI and enterprise security are actually headed.

“I’ve had many conversations with him about where this is going, and every one of those conversations is extremely provocative,” Hemant told me, “and so it made sense to lean on him. Just like we lean on Ken.”

Tech Lash 2.0 & The 4 Forces Driving It

The review’s most pointed section is Hemant’s argument that “Tech Lash 2.0” is already building around AI, and that it’s largely self-inflicted. He cites a Wall Street Journal piece calling this “The Most Joyless Tech Revolution Ever” and agrees with the framing. The anxiety spreading across society about displacement, erosion of human connection, and what intelligent machines mean for ordinary lives is, in his view, fully rational, and largely a response to who Silicon Valley has become in public. He identifies four forces that have brought us here.

The first is the social-media optimization of our minds. Hemant argues the industry has given itself false permission to decouple its social presence from its real-life presence. “Moderation does not get engagement,” he writes. “Nuance does not go viral. And so we have collectively abandoned both.” When I asked him about it directly, he was blunt: saying ridiculous things for attention is “a entirely unproductive exercise if you’re focused on the long term. It is a very short term mindset.”
The second is politics contamination. When Silicon Valley entered the political arena in recent years, Hemant argues it had the opportunity to export its best qualities, bias for action, intellectual honesty, genuine meritocracy. Instead, it imported the worst qualities of politics: tribal sorting, ideological litmus tests, zero-sum conflict performed for audiences rather than resolved for outcomes. GC’s response was to decline political engagement and instead launch the GC Institute, investing in policy education across DC, Brussels, and Delhi. “We had the institutional credibility to elevate public discourse,” Hemant writes. “We degraded it, and our own industry, instead.”
The third is the genius asshole myth. Hemant’s argument here is empirical: organizational behavior research consistently shows that one genuinely toxic person costs an organization hundreds of thousands annually in lost productivity, turnover, and diminished performance. When someone succeeds and happens to be cruel, we assume causation when the evidence shows coincidence. “They succeed despite it,” he writes. “They would succeed more without it.” The theory that brilliance licenses cruelty, in his view, is both statistically wrong and a license most people invoking it don’t qualify for.
The fourth, and the one Hemant has made personal mission to counteract, is contempt for legacy. In Silicon Valley’s culture, “legacy” has become a synonym for irrelevance. He considers this precisely backward. “Legacy is a beautiful word and a beautiful thing to have built. The word itself tells you what matters: you endured. You were trusted long enough, by enough people, to leave a mark.” The industries tech most wants to transform are governed and trusted by people with decades of institutional knowledge the industry doesn’t have. Treating them as obstacles rather than partners is, in his view, how the transformation stalls.

The Case for Kindness (Not Niceness)

When I asked Hemant about the kindness argument in our interview, he was careful to draw a distinction that matters. Kind is not the same as nice. “Kindness and ambition are not at odds with each other,” he told me. “I think in Silicon Valley, we try to glorify the asshole symptom of founders thinking that’s almost a necessary ingredient to succeed. And I don’t think it has to be that way.”

Niceness, in his framing, is often its own form of contempt, conflict-avoidant, performative, unwilling to engage seriously with ideas that challenge. Kindness is different. Patrick Collison delivers difficult feedback and disagrees directly. Satya Nadella inherited a Microsoft defined by stack-ranking and internal warfare and disbanded the toxic culture while raising expectations. Paul Graham described his purpose at YC as “trying hard to prevent assholes from getting funded.” Roger Federer was the most ferociously competitive player of his generation and never once diminished an opponent.

Hemant ties this directly back to GC’s investment thesis. Transforming healthcare, defense, energy, and education at the scale the firm envisions requires trust from governments, regulators, legacy institutions, and communities who have no obligation to extend it. Burning that trust for engagement metrics and personal brand isn’t just a cultural issue, it’s a business risk that compounds across every deal GC wants to do. His closing line captures the argument cleanly: “AI can’t cure cancer if we’ve convinced everyone it is a cancer.”

The quarterly review, in that sense, is itself part of the answer: slower, more honest, more willing to sit in the gray. It’s a bet that the industry’s most durable competitive advantage over the next decade won’t be capability. It will be whether we were worth rooting for.

Read the Q1 2026 Review

→ Listen on X, Spotify, YouTube, Apple

Subscribe now

ICYMI: Sourcery 🤝 NYSE

Sat, 18 Apr 2026 14:29:28 GMT

Brought to you by Brex:

Brex, the ultra-superintelligent finance platform for elite businesses, combining the world’s smartest corporate card with integrated expense management, banking, bill pay, & travel. Trusted by 35,000+ companies, including Anthropic, OpenAI, Arm, Robinhood, ServiceTitan, DoorDash, & Wiz. Built for scale. Spend smarter, move faster.

Sourcery 🤝 NYSE

Some personal news: Sourcery is officially partnering with the NYSE. 🔔

Sourcery & NYSE are partnering to keep the “brrr” in breaking IPOs, tech financings, & markets coverage — interviews with founders, CEOs, & investors.

My goal with Sourcery has always been to have investing & tech conversations in an authentic, informative, & fun way (jury’s still out on if the Jack Altman/Sam Altman clip is real). Getting to do that from the floor of the New York Stock Exchange with the support of their team is next level. I couldn't have asked for a better start to 2026.

This partnership means Sourcery will be filming & covering select NYSE listings, market events, & milestones throughout the year, plus collaborating on more interviews ( Lynn, if you’re reading this - hot take? ). Expect more IPO days, CEO sit-downs, investor conversations, & the occasional hot take. Bi-coastal Sourcery is about to be in full swing.

Huge thank you to Lynn Martin, Joe Benarroch, Kevin Hawkins, Lance Glinn, & Kristen Scholer for believing in what we’re building & making room for an independent & fast growing voice on the floor. I couldn’t be more excited.

Very special shoutout to Christian for the intro wayyyy way back last fall.

PS — If you’re a very large space or AI company planning a HUGE IPO.. or a CEO or investor with an amazing story, the NYSE floor is waiting. Reply to this email or DM me on X.

PPS — And yes, rumors are circulating this may have been bigger than the Polymarket deal 😂

PPPS — If you have any leads on any of the big names below, would love any help in making these interview happen! Thank you :)

ICYMI: Public @ NYSE

Leif Abraham, Co-CEO of Public, joins Sourcery at the New York Stock Exchange for one of the more fun discussions on finance & how Public is building the next-gen investing platform “for those who take it seriously.” Public represents the first modern alternative to Schwab, Fidelity, and Vanguard — platforms built on decades-old infrastructure that haven’t kept pace with today’s investors or technology.

→ Listen on X, Spotify, YouTube, Apple

Subscribe now

Brought to you by:

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Merge—The leading provider of customer-facing integrations and agentic tools for frontier LLMs, Fortune 500 organizations, and B2B SaaS companies. Visit: https://merge.dev

NYSE 🤝 Sourcery Interviews

→ Listen on X, Spotify, YouTube, Apple

🔔 NYSE-Listed Companies

→ Listen on X, Spotify, YouTube, Apple

@NYSE","username":"MollySOShea","name":"Molly O’Shea","profile_image_url":"https://pbs.substack.com/profile_images/1673061505351450624/2A8YvQ2N_normal.jpg","date":"2026-04-06T19:01:16.000Z","photos":[],"quoted_tweet":{"full_text":"Bright lights, big tech. \n\n@sourceryy X NYSE\n\nWatch this space, with @MollySOShea","username":"NYSE","name":"NYSE 🏛","profile_image_url":"https://pbs.substack.com/profile_images/1505733102148145153/w2gDOam4_normal.jpg"},"reply_count":134,"retweet_count":38,"like_count":1025,"impression_count":147729,"expanded_url":null,"video_url":null,"belowTheFold":true}" data-component-name="Twitter2ToDOM">

@sourceryy X NYSE\n\nWatch this space, with @MollySOShea ","username":"NYSE","name":"NYSE 🏛","profile_image_url":"https://pbs.substack.com/profile_images/1505733102148145153/w2gDOam4_normal.jpg","date":"2026-04-06T18:41:39.000Z","photos":[{"img_url":"https://substackcdn.com/image/upload/w_1028,c_limit,q_auto:best/l_twitter_play_button_rvaygk,w_88/mnxtpdavzslg8xj3oxmb","link_url":"https://t.co/6Bub9gPV07"}],"quoted_tweet":{},"reply_count":37,"retweet_count":25,"like_count":349,"impression_count":354576,"expanded_url":null,"video_url":"https://video.twimg.com/amplify_video/2041224517142597632/vid/avc1/720x1000/-E2AvwscXZ_7rpTs.mp4","belowTheFold":true}" data-component-name="Twitter2ToDOM">

Subscribe now

BREAKING: The AI Brokerage Quietly Taking on Schwab, Fidelity & Vanguard

Molly O’Shea — Fri, 17 Apr 2026 13:58:22 GMT

The Great Wealth Transfer

→ Listen on X, Spotify, YouTube, Apple

The results speak for themselves: millions of users, billions in assets under management, billions in trading volume, and revenue doubling annually as they continue to capture market share.

Public’s differentiators include:

AI agents that automate complex investment strategies
AI-powered custom indices with “Generated Assets”
Comprehensive multi-asset trading across stocks, ETFs, crypto, options, treasuries, bonds
Sophisticated investors with real assets to compound

Leif and I explore how their AI-generated defense index hit 350% returns in one month (big surprise.. lol), the future of financial advisors in an AI world, and why their co-CEO structure works better than solo leadership. Leif explains why they focus on the top 25% instead of chasing mass market speculation, plus his thoughts on AI, wealth inequality, and what it means for society.

For investors seeking a serious platform that combines traditional financial services with cutting-edge technology, this conversation offers valuable insights into where the industry is already aimaxxxxing & what comes next.

Subscribe now

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

(00:00) Leif Abraham, Co-Founder & Co-CEO at Public
(01:10) Competing with traditional brokers
(03:45) AI agents for investing
(05:22) Are Public’s videos AI or real?
(06:31) Why the co-CEO model works
(09:34) The evolution of Public
(15:06) Growth, scale, and market cycles
(17:59) Breaking down the product ecosystem
(20:45) Public's design philosophy
(22:59) Building portfolios with AI
(26:33) Ideal customers
(30:20) The decline of financial advisors
(33:00) What AI replaces in wealth management
(35:45) Risks and limits of AI agents
(37:32) The “financial super app” debate
(42:39) Prediction markets vs real investing
(45:04) Will everything become tradable?
(46:22) Tokenization: the hype vs. the reality
(50:37) The future of financial platforms
(55:56) VCX, closed-end funds & speculation risk
(59:09) Would Public go public?
(01:02:22) What’s next for Public
(01:04:08) The dark AI prediction
(01:06:52) Wealth taxes vs. UBI

Brought to you by:

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Merge—The leading provider of customer-facing integrations and agentic tools for frontier LLMs, Fortune 500 organizations, and B2B SaaS companies. Visit: https://merge.dev

Public’s AI-First Challenge to Wall Street’s Old Guard

How Leif Abraham Is Building the Next-Generation Brokerage That’s Stealing Accounts from Schwab and Fidelity

In a candid interview at the New York Stock Exchange, Public Co-CEO Leif Abraham revealed how his company is quietly disrupting the investment management industry.. not by chasing retail day traders, but by building a premium platform designed for serious investors in the AI era.

With $400 million raised across five funding rounds and a product suite spanning stocks, bonds, crypto, options, & innovative AI-powered tools, Public represents the first credible challenge to legacy giants like Charles Schwab, Fidelity, and Vanguard in decades. With millions of users, billions in assets under management, and revenue doubling annually.. they’re well on their way.

→ Listen on X, Spotify, YouTube, Apple

The Strategic Pivot: Targeting the Top 25%

Unlike most fintech startups that chase mass market adoption, Public made a deliberate strategic decision in 2022 to focus exclusively on affluent investors — what Abraham calls “the top quartile.”

“In the US, sadly, if you are in the sub 75%, you live paycheck to paycheck, credit card to credit card, you do not have money to compound in the markets. You just don’t,” Abraham explained. “And that’s also why you see the people who acquire those & build products for those [customers], and consumer FinTech, you have to monetize those through loan speculation or interchange or something.”

This “Level Up” strategy fundamentally changed Public’s approach. Rather than building a simplified app for mass adoption, they created a sophisticated platform for investors with real assets to compound. The competitive results are proving it out: Public’s ACAT (Automated Customer Account Transfer) data shows they’re winning established investors directly from legacy incumbents.

“Actually, if you look at where people ACAT accounts from, where they like consolidate brokerage accounts from, for us it’s Schwab, fidelity, E-Trade, Vanguard, and then Merrill, and Robinhood.”

This metric is particularly significant because ACAT transfers represent existing investors with established portfolios choosing to move their entire accounts — stocks, bonds, ETFs, and cash — from their current brokerage to Public. Unlike new account openings, these transfers require convincing someone that Public is superior enough to their current platform to justify the effort of moving. The fact that traditional giants like Schwab, Fidelity, and Vanguard top the list, with newer platforms like Robinhood trailing, demonstrates Public is genuinely disrupting incumbents rather than just competing for first-time investors.

“We’re not gonna build the money super app. We are rather gonna be the investing super app for lack of a better wording,” Abraham said, explaining why they reject the “super app” trend sweeping fintech.

Funding Journey: Social Investing to AI Unicorn

Public’s $400 million funding trajectory mirrors both the retail trading boom and the company’s strategic evolution toward AI-powered investing. The journey began modestly with seed funding of $2 million in 2017-2018, followed by a $9 million Series A in 2019 to establish their initial social investing product.

Momentum accelerated with a $15 million Series B in March 2020, led by Accel and Greycroft, featuring high-profile strategic investors including Will Smith (via Dreamers VC), Sophia Amoruso, and NFL star J.J. Watt. The timing proved fortuitous as retail trading exploded during the pandemic.

Public’s breakout moment came with a $65 million Series C in December 2020, again led by Accel alongside Lakestar and Greycroft, with additional participation from cultural investors like Tony Hawk and the Chainsmokers’ Mantis VC. This was quickly followed by the company’s unicorn round: a massive $220 million Series D in February 2021 at a ~$1.2 billion valuation, with backing from Accel, Greycroft, Tiger Global Management, and Inspired Capital.

After a quiet period during the 2021-2022 market reset, Public resumed fundraising with a $135 million Series D-2 round in December 2024, including approximately $105 million in equity and $30 million in debt. Accel’s continued leadership across multiple rounds signals sustained conviction from their earliest institutional backer.

The funding composition reflects Public’s positioning at the intersection of consumer fintech, media, and next-generation market infrastructure, consistently attracting both top-tier venture firms and culturally relevant strategic investors.

AI Agents: Future of Investment Management

Public’s most ambitious product launch centers on AI agents that automate complex investment strategies. Unlike simple robo-advisors, these agents can execute sophisticated workflows based on natural language instructions.

“What that helps you do is to essentially automate any type of workflow in your brokerage account. That could be a trading strategy. There can be money movements, that can be alerts you wanna set up,”
“It really moves the way you even just like interact with the platform away from just like clicking buttons to just expressing intent.”

The safety mechanism separates reasoning from execution — users collaborate with AI to design strategies, but the final implementation runs deterministically without hallucination risk. “We essentially separate the reasoning from the execution... once you approve it, it is a deterministic model where that can’t deviate,” Abraham noted.

This approach has enabled sophisticated strategies like covered call income generation, where users can simply ask: “Look at my portfolio. Show me if I could do, you know, five grand a month an income potential on it.”

Generated Assets: Custom Indices Through AI

Perhaps Public’s most innovative product, Generated Assets allows users to create custom investment indices through natural language prompts. The AI scans the entire US stock universe, evaluates companies against user criteria, and builds a tradeable index complete with backtesting data.

One particularly successful example caught attention: “I think the best performing generated asset on the platform right now is defense and weapons manufacturers. That is up ~350% in the last month.”

Beyond thematic investing, users leverage the tool for sophisticated portfolio construction. “Give me, you know, the NASDAQ 100, but truly take out anything that’s not tech,” Abraham cited as a common use case, allowing investors to remix existing indices with precision unavailable through traditional ETFs.

The Co-CEO Advantage

Public’s dual-leadership structure with Co-CEO Jannick Malling reflects European business practices but remains unusual in US venture capital. Abraham argues this creates better decision-making through balanced perspectives.

“I think the biggest thing is, I think first it helps really for more balanced decision making,” Abraham explained. “We’re all human to some regard. We’ll all have a time where we go to the office and we’ll feel anxiety or we’ll be in a bad mood or whatever... And it helps to have someone else who experiences everything the same way you do.”

The structure divides responsibilities clearly: Malling oversees engineering and product design, while Abraham leads growth and business development. This allows each co-CEO to develop deep expertise while maintaining shared accountability for major decisions.

The End of Financial Advisors

Abraham predicts AI will systematically disrupt the traditional financial advisory model by replacing three core functions: grunt work, advice, and emotional support.

“If we look at what a financial advisor does for you, we break down these in these three things where it’s like,
#1 grunt work. So like doing the actual tax loss harvesting, setting up some strategy for you, etc.
#2 is advice, which is basically opinions on the back of news, data, reports
#3 is emotional support.”

AI agents already handle the first category. Abraham believes emotional support may actually be easier to replace than most assume: “Considering you have people having full relationships with AI already now... I think that part is actually easy to replicate.”

The advisory industry faces demographic headwinds that compound the AI threat.

“There’s less people becoming financial advisors now. New generations. The average age of financial advisor that goes up by year every year. I think the average age is actually roughly 60 or so already.”

Contrarian Views on Prediction Markets and Tokenization

While billions flow into prediction markets and tokenization projects, Abraham takes a measured skeptical stance on both trends.

On prediction markets: “I think prediction markets or event contracts... are very powerful tool. We’re gonna go into that somewhat ourselves, but I think right now, the use of events, contracts or the use cases are being stretched.”

He sees value in securities-focused contracts for sophisticated portfolio management but criticizes the expansion into entertainment betting: “Using it to essentially do sports betting... I would argue it’s kind of, you know, stretching the use case.”

Regarding tokenization, Abraham cuts through the hype: “Just because you wrap it in a token doesn’t mean it’s trading 24/7... it’s a fucking wrapper. The issue is liquidity. It’s not if there’s a wrapper of a token.”

The Dark Side: AI & Wealth Inequality

Despite his enthusiasm for AI’s potential, Abraham harbors serious concerns about societal implications. He predicts a difficult transition period before any utopian AI future.. hot take incoming..

“Before we live in this oasis where no one has to, where, where no one has to work anymore and whatnot... I think it’s gonna get really dark and it’s gonna get really dark because... The people that have access to the tools that can use them, that own the tools... the power disparity that’s gonna happen there is gonna be so massive.”

Abraham envisions stark societal changes: “New York is gonna feel more like São Paulo than it does like New York right now, I think... in your New York fancy apartment where you have your doorman, it’s not gonna be a doorman, it’s gonna be an armed security guard.”

The solution, he believes, could require both universal basic income and purpose creation: “It’s not just about creating UBI, it’s also about what can we, can we create that, uh, contingency gives people purpose.”

→ Listen on X, Spotify, YouTube, Apple

About Public: Founded in 2017 by Leif Abraham and Jannick Malling, Public has raised approximately $400 million across five funding rounds, with backing from Accel, Tiger Global, Greycroft, and Lakestar. The platform offers comprehensive multi-asset trading, AI-powered investment tools, and serves millions of users with billions in assets under management. Public’s tagline “Investing for those who take it seriously” reflects their premium positioning against legacy incumbents like Schwab, Fidelity, and Vanguard.

Subscribe now

NEW: Giant Magnifying Glass + Hot Rocks Powering AI Data Centers

Thu, 16 Apr 2026 13:50:40 GMT

Giant Magnifying Glass + Heated Rocks

Most announced data centers will never get built. The ones that do are running on gas generators. And the US grid has no answer for what's coming.

→ Listen on X, Spotify, YouTube, Apple

Hannan Happi, CEO & Co-Founder of Exowatt, joins Sourcery for a rare facility tour of their 40,000 sq ft Miami HQ — walking us through the tech, the team, and why the AI power crisis is far worse than the headlines suggest.

Exowatt has raised $140M in total funding from a16z, Atomic, Felicis, Sam Altman, Leonardo DiCaprio, Starwood Capital, Thrive Capital, & more, all betting on a deceptively simple idea: concentrate sunlight with Fresnel lenses, store the heat in rocks at 1,000°C, and dispatch electricity on demand. 24 hours a day, no grid required.

Their P3 system is built from sand, dirt, and steel. Like a large magnified glass and a rock. No lithium. No cobalt. No China. Target cost: 1 cent per kilowatt hour.

In this episode:

→ Why most announced data centers are phantom projects
→ The real bottleneck that replaced GPUs: power & skilled labor
→ How a 200-year-old engine is at the heart of their stack
→ What hyperscalers actually say about sustainability (it's not pretty)
→ Lessons from Tesla on iteration, modularity & vertical integration
→ Why Miami — and why now

With a 90+ GWh demand backlog and commercial deployments live in 2025, this might be the most important energy company you haven't heard of.

Co-Founded in 2023 by Hannan Happi & Atomic CEO Jack Abraham, Exowatt’s mission is to make sustainable renewable energy always available & almost free.

Subscribe now

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

(00:00) What is Exowatt? The solar backbone for AI data centers
(01:30) Why the US grid simply cannot handle what's coming
(02:00) Tour begins: Welcome to Lighthouse Miami — 40,000 sq ft facility
(02:45) The P3 explained: three elements, one shipping container
(04:00) The founding story: Atomic venture studio and the modular hypothesis
(05:30) Why solar thermal hasn't followed solar PV's 99.6% cost decline — until now
(07:00) North star: 1 cent per kilowatt hour and where they are today
(11:00) Inside the world's largest solar simulator — the sun, indoors
(13:00) The Fresnel lens up close: a giant magnifying glass for the sun
(15:00) "It's rock science" — the 5-year-old explanation of how it works
(17:30) Domestic raw materials (sand, dirt, steel) as the real competitive moat
(24:30) The phantom data center problem: most announcements won't get built
(26:00) From 100 MW to 10 GW: how data center scale exploded in 3 years
(30:00) Funding breakdown: $140M, a16z, Felicis, Sam Altman, Leonardo DiCaprio
(37:00) Lessons from Tesla: remove parts, iterate fast, build better

Brought to you by:

Turing—Turing delivers top-tier talent, data, and tools to help AI labs improve model performance, and enable enterprises to turn those models into powerful, production-ready systems. Visit: turing.com/sourcery

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Merge—The leading provider of customer-facing integrations and agentic tools for frontier LLMs, Fortune 500 organizations, and B2B SaaS companies. Visit: https://merge.dev

The AI Power Crisis Is Bigger Than Anyone Is Admitting

Inside Exowatt’s bet that superheated rocks and a 200-year-old engine can power the next wave of AI infrastructure.

→ Listen on X, Spotify, YouTube, Apple

The Phantom Data Center Bubble

There is a version of the AI infrastructure story that gets told in press releases, keynote slides, and analyst reports. In that version, hyperscalers are constructing a grid of colossal data centers across America, a trillion-dollar buildout that will define the next era of computing. It is a compelling story. The reality on the ground looks considerably different.

Hannan Happi, CEO and Co-Founder of Exowatt, works directly with the hyperscalers building these facilities. His read is straightforward: a lot of the data centers being talked about will never get built.

“There’s a lot of headlines about data center projects being planned or initiated.. I don’t think all of them are actually happening. So there’s a lot of like phantom data centers that are being announced.”

The evidence for this is not hard to find. Stargate, the $500 billion AI infrastructure initiative announced at the White House in January 2025 by OpenAI, Oracle, and SoftBank, was positioned as the defining infrastructure bet of the AI era. Within six months, Bloomberg reported that the joint venture had not hired any staff and was not actively developing any data centers. The three partners had been arguing over structure, financing, and site selection since the day it was announced. OpenAI, unable to secure financing for its own data centers, eventually returned to Oracle and SoftBank for a restructured two-way deal. The Abilene, Texas expansion, which was meant to grow from 1.2 gigawatts to 2 gigawatts, was subsequently scrapped. OpenAI also paused its Stargate UK project entirely, citing energy costs and regulatory uncertainty.

Microsoft told the market in January 2025 that it would spend $80 billion on AI data centers that year. By March, TD Cowen analysts reported that Microsoft had walked away from more than 2 gigawatts of data center leases across the US and Europe, canceled a $1 billion campus development in Ohio across three sites, and frozen new leasing activity almost entirely. The company maintained it was reallocating rather than retreating, but the gap between the January announcement and the March reality was significant.

This pattern, large announcements followed by quiet pullbacks, is exactly what Happi is describing. Land acquisition, permitting, grid interconnection, financing complexity, and the availability of skilled labor all create friction between a press release and a functioning data center. The gap between announced capacity and operational capacity is measured in years and gigawatts, and it has become one of the more important blind spots in the AI conversation.

The Underlying Demand Is Real. That Is Where Exowatt Comes In.

Calling out the phantom project problem is not the same as saying the demand for AI power is overstated. It is not. The International Energy Agency projects that US data center power consumption is on course to account for almost half of the growth in electricity demand between now and 2030. That is not a bubble. That is a structural shift in how the country uses energy, and it is happening regardless of which specific projects get built on which specific timelines.

The distinction Happi is drawing is between announced capacity and real demand, and Exowatt’s own commercial traction illustrates it clearly. When the P3 launched at RE+ 2024, North America’s largest clean energy conference, the company reported a pipeline of more than 1.2 gigawatts of orders. That grew to 85 gigawatt hours of signed customer demand by the end of 2024, and exceeded 90 gigawatt hours by April 2025. Those are not letters of intent or press release projections. They are contracts from data centers, energy developers, & hyperscalers who need power.. & need it now.

The speed at which that backlog has grown reflects something specific: customers are not evaluating Exowatt as a 10-year bet. They are evaluating it as a solution to an immediate problem. The P3 can be deployed with or without grid interconnection, ships in standard 40-foot containers, requires minimal on-site construction, and can be co-located directly next to a data center facility. For a hyperscaler staring down a five-to-seven year grid interconnection queue, a modular system that generates power on-site from day one is not a nice-to-have. It is the only durable option that fits the timeline.

Exowatt has since expanded beyond selling hardware. In January 2026, the company launched ExoRise, a new business arm that delivers fully powered land and energy infrastructure for hyperscale data centers across the US Southwest, covering New Mexico, West Texas, Arizona, and Nevada. The pitch is a turnkey alternative to years of utility bureaucracy: show up, plug in compute, & Exowatt handles everything else.

“Hyperscalers should not have to piece together land, power, & infrastructure on their own,” Happi said at the launch. “ExoRise delivers everything together so customers can focus on compute while we handle the power.”

The demand is real. The grid cannot meet it. And Exowatt is already shipping.

Power Is the Known Bottleneck. Here Is What That Actually Means.

The power constraint in AI infrastructure is no longer a hidden insight. It has been covered extensively, referenced in every major hyperscaler earnings call, and cited as the primary reason behind dozens of infrastructure decisions in the past year. What is less well understood is the specific mechanism by which it constrains the buildout, and why throwing money at it does not solve it quickly.

A single gigawatt data center, the kind that was considered large just three years ago, consumes the equivalent energy of nearly one million US households. Hyperscalers now planning five and ten gigawatt campuses are effectively trying to plug small cities into infrastructure designed for suburbs.

“The grid is just really not built for that,” Happi said. “And what that means is the data centers now have to figure out how to bring their own power online.”

The specific mechanism is the interconnection queue, the process by which new power generation projects formally connect to the US grid. That queue is backed up for years in most markets. A developer who applies today for grid interconnection in a high-demand region can expect to wait five to seven years for approval and connection. No amount of capital accelerates that timeline in any meaningful way because the constraint is regulatory and physical, not financial. Transmission infrastructure capable of serving a ten-gigawatt facility in a rural location does not exist and cannot be built on the timelines hyperscalers need.

The outcome is a shift that would have been considered unusual just two years ago. Data centers are going off-grid. Behind-the-meter power generation, building your own power source on-site rather than drawing from the utility, has moved from an edge case to a near-standard consideration for any large-scale build. The question hyperscalers are now asking is not whether to generate their own power. It is what to generate it from, and how fast they can get it online. The answer most are currently landing on is natural gas, with all of the cost, environmental, and geopolitical consequences that come with it.

Big Tech Quietly Moved On From Clean Energy

For most of the past decade, the major technology companies competed on sustainability as seriously as they competed on product. Net-zero commitments, renewable energy certificates, carbon neutrality pledges. Environmental goals were a fixture of earnings calls and annual reports. That has shifted, and the shift has happened quickly.

Happi describes the current priority order inside hyperscaler conversations plainly. Power availability comes first, at almost any cost and from almost any source. Cost comes second. Environmental considerations, which once sat near the top of the agenda, have fallen considerably down the list, behind reliability, speed of deployment, geopolitical risk, and community impact.

“It used to be one of the objectives,” he said of sustainability. “But it’s now time to power, right? I need power. I don’t care how you make it. Squeeze penguins if you have to, I need the power.”

The result is that natural gas has become the default power source for off-grid data center development. Gas generators are available, scalable, and well-understood. They are also expensive over time, carbon-intensive, and in an environment where energy security has become a national priority, strategically exposed.

The longer-term risk is lock-in. Gas infrastructure built today to power a data center will run for twenty years. The clean energy commitments set aside to enable that build do not go away. They accumulate. The costs associated with emissions, regulatory pressure, and rising energy prices will arrive eventually. The current thinking inside the industry is that speed to deployment outweighs those concerns for now.

Why the Answer Might Be Sand, Dirt, & Rocks

Exowatt’s technology is a direct response to this situation. The Exowatt P3 is a modular, shipping container-sized power generation unit built on straightforward physics and a domestic material supply chain.

The system works in three stages. A Fresnel lens, the same optical technology used in lighthouses, concentrates incoming solar energy onto a thermal battery made of specialized rocks. Those rocks are heated to between 800 and 1,000 degrees Celsius, storing energy for up to five days without additional solar input. That heat is then moved via a circulating air loop to a Stirling engine, a technology that has been in use for 200 years, which converts the temperature differential into electricity on demand.

None of the individual components are new. Fresnel lenses, thermal storage, and Stirling engines are all well-established technologies. The value is in combining them in a way that is modular, factory-producible, and built entirely from domestic materials: sand, dirt, and steel, with no lithium, cobalt, or rare earth minerals involved.

“The most important thing is at the end of the day the raw materials that we use: sand, dirt & steel,” Happi said. “And so you don’t have to go import something from China.”

Exowatt’s current cost is just under four cents per kilowatt hour (!!). Their long-term target is one cent, which would make it among the cheapest electricity ever produced at commercial scale. Getting there depends on manufacturing volume, which is the core thesis behind how the company was designed.

Economics of Power & Why Exowatt is Ready Now, Not in 10 Years

Hyperscalers signing power contracts today are paying in the range of 20 to 30 cents per kilowatt hour. Commercial grid electricity typically runs between 8 and 15 cents. Residential customers pay 13 to 20 cents. The premium data centers are absorbing reflects a straightforward imbalance: there is not enough power available where they need it, and they will pay whatever it takes to secure what exists.

Natural gas, the current default for off-grid generation, is not cheap either. Once you factor in fuel costs, transportation, generator maintenance, and long-run exposure to gas price volatility and carbon penalties, the economics of running a gigawatt data center on gas for twenty years are significantly worse than the upfront simplicity of the decision implies.

Exowatt’s current cost of just under four cents ($0.04) per kilowatt hour already undercuts the grid price for commercial customers in most US markets. At scale, with the manufacturing learning curve applied, the one cent ($0.01) target would make it cheaper than any grid-connected power source available today. That is the economic destination. The more immediate question is whether the product works on the timeline the market actually needs.

This is where the modular, container-based approach changes the conversation. A traditional utility-scale power project, whether solar farm, gas plant, or nuclear, requires years of permitting, grid interconnection queues of five to seven years in most high-demand markets, and construction timelines that assume a patient developer with a long capital runway. None of those conditions describe the AI infrastructure buildout happening right now.

The P3 sidesteps all of it. Because each unit is factory-built and ships in a standard 40-foot container, it can be deployed on-site without grid interconnection. A data center that needs power in 18 months does not need to wait for a utility. It needs land, adequate solar irradiance, and P3 units. The US Southwest, where Exowatt is targeting its ExoRise deployments, has all three in abundance.

The comparison to competing approaches is stark.

Utility grid connection in a constrained market: ~5-7 years.
New gas plant permitted & built from scratch: ~3-5 years
Nuclear small modular reactor, which several hyperscalers have explored as a long-term solution: ~10 years or more
- Despite a lot of support, they’re currently in an early, pre-commercial stage
Exowatt’s P3, deployed in modular increments on pre-permitted land: online within the current build cycle

Customers are not just buying the one cent per kilowatt hour target. They are buying the ability to get power online on the timeline that AI actually demands, from a domestic supply chain that does not depend on a utility, a foreign mineral, or a gas contract that locks in costs for two decades.. or pose a major public health hazard.

The Bottleneck After Power: Skilled Labor

The AI infrastructure conversation has worked through a clear sequence of constraints. First it was Nvidia GPUs. Then, as chip availability improved, attention moved to power. Both received significant coverage and drove substantial capital toward solutions.

There is a 3rd bottleneck developing that has received far less attention: skilled labor.

“The biggest kind of challenge for data centers today is there aren’t enough construction workers to build a data center,” Happi said. “You have to import people from other states to a site to build a data center. And being an electrician these days is actually a very high-paying job because there aren’t enough electricians to even twist wires and hook things up.”

Unlike the chip shortage, this one does not respond quickly to capital investment. Training skilled tradespeople takes years, and the industry did not begin investing in that pipeline early enough to meet current demand. The result is a constraint that will slow the buildout well after the power availability problem is addressed.

Exowatt’s modular approach has a practical advantage here. Units assembled in a controlled manufacturing environment and shipped to site require less on-site skilled labor than bespoke, utility-scale power infrastructure. When qualified electricians are the limiting resource, reducing how many hours of their time each megawatt requires is a meaningful efficiency gain.

Exowatt Careers

The Manufacturing Bet Behind the Technology

The core idea behind Exowatt is not any single piece of technology. It is a manufacturing thesis drawn from two reference points: the cost trajectory of solar PV panels over the past fifty years, and the production philosophy Happi observed while working at Tesla.

Solar PV reduced its cost per watt by 99.6% over five decades. That did not happen because of one breakthrough. It happened because manufacturing scale drives learning, learning drives efficiency, efficiency drives lower cost per unit, and lower cost per unit enables more volume. The curve compounds.

Solar thermal, the technology category Exowatt operates in, has not followed that trajectory. Every solar thermal project built to date has been a large, custom installation. There is no learning curve when every project is built differently. Cost has stayed high because the manufacturing discipline that transformed solar PV was never applied.

Exowatt’s approach is to apply it. Shipping container-sized modules, built in a factory to a consistent specification, create the conditions for that learning curve to develop. Each generation of the P3 informs the next. The one cent per kilowatt hour target is the projected outcome of running that process long enough and at sufficient volume.

“That’s why we chose this modular approach,” Happi said. “Instead of going & building a giant infrastructure project and saying we need ten years and a billion dollars to get to our first demo site, we said: let’s build the smallest thing, learn from that, iterate quickly, & build the next generation, & the next generation.”

Backed by $140 million from a16z, Felicis, Sam Altman, Leonardo DiCaprio, Starwood Capital, Thrive Capital, and others, with a demand backlog exceeding 90 gigawatt hours from data centers and hyperscalers across the US, Exowatt is moving from thesis to deployment. Several commercial projects are going live in 2025. The manufacturing scale that makes the cost curve work is what comes next.

Hannan Happi & What Tesla Taught Him

Before Exowatt, before Atomic, before Miami, & before founding Volansi, Hannan Happi was on the factory floor at Tesla during the launch of the Model S, when the Fremont factory was being stood up from scratch.

“Tesla wasn’t a popular company. I had to fight a lot, uphill, to get there. But it was an amazing, eye-opening experience.”

That experience left a set of convictions that run through everything Exowatt does, and Happi is direct about what they are.

1.) Team >

The team is the primary ingredient. Not the tech, market timing, or capital.

“The lessons learned were to first of all have an amazing, dedicated, mission-focused team. I think that’s the primary ingredient that actually got Tesla through.”

The reason Tesla survived, in Happi’s view, was a group of people willing to absorb a level of pressure that would dissolve a more conventional organization. Exowatt was built with that standard in mind, which is part of why the company chose Miami deliberately rather than defaulting to the Bay Area talent pool, and why the 40,000 square foot facility in Magic City looks more like an aerospace manufacturing floor than a startup office.

2.) Build What People Want

The second lesson Happi traces directly to his time at Tesla and his time at YC: build something people actually want.

“Build a great product that customers really want and enjoy. That’s something I also learned at YC, right from Sam. Build something people want.”

3.) Remove Parts.

The instinct in engineering is to add capability. The discipline that separates companies that scale from companies that stall is the willingness to take things away.

“That constant iteration on: let’s remove parts, let’s make it simpler, let’s scale the manufacturing — that’s the same approach you see at SpaceX or at Tesla. That’s what has led to this.”

Every configuration of the P3 was evaluated against the bill of materials. Every component that could be eliminated was eliminated. The result is a system whose simplicity is what makes it manufacturable at scale.

4.) Short Feedback Loops

Hardware is hard, as Happi puts it, not because the physics are complicated but because iteration cycles are long and expensive if you let them be.

“Hardware is hard. So you want to try to figure out how to create the shortest feedback loops possible & learn from that as fast as possible, & scale that as fast as possible. That’s why we chose this modular approach.”

Exowatt went through more than 50 configurations of the P3 before landing on the version that went to market. The modular approach was not just a product decision. It was a way of compressing the time between building something, learning from it, and building the next version.

5.) Vertical Integration

Or at least the ambition toward it. Happi is candid that Exowatt has not gone as far down that path as Tesla yet.

“Everyone coming out of Tesla is obsessed with vertical integration. You want to build everything in-house. We’ve started working with contract manufacturers because we know we need to scale fast. But Elon really taught us how to vertically integrate everything & build it yourself, build it better, faster, cheaper, & don’t take no for an answer.”

The companies that win in hardware over the long run are the ones that control their own supply chain, their own manufacturing process, and their own cost structure. Every move Exowatt makes toward domestic sourcing and factory production is a step in that direction. The Tesla years gave Happi a clear picture of what that looks like at full maturity. Exowatt is working backward from it.

The Investors & Origin Story Behind It All

Exowatt did not come out of a garage. It came out of Atomic, the San Francisco-based venture studio co-founded by Jack Abraham, who also serves as Exowatt’s Chairman. Atomic’s model is to build companies from the inside, pairing founders with capital, infrastructure, and operational support before a company has a product or a customer. It has produced a number of significant companies including Hims & Hers Health and OpenStore. Happi eventually “moved in as a founder in residence back in 2022 & started working with Jack” ultimately, incubating Exowatt & formally Co-Founding the company with Abraham in 2023.

The founding thesis was not “build a solar company.” It was more specific than that: apply manufacturing discipline to solar thermal in the same way the industry applied it to solar PV, do it at the modular scale, & source domestically. That specificity is part of what attracted the caliber of investors who came in at the seed round.

In April 2024, Exowatt came out of stealth with a $20 million seed round led by Andreessen Horowitz and Atomic, with Sam Altman, CEO of OpenAI, and Leonardo DiCaprio investing as angels. The combination of Altman & DiCaprio in the same cap table is unusual enough to earn attention on its own, but the more meaningful signal was a16z leading the round. Andreessen Horowitz has been one of the more disciplined investors in energy infrastructure, and their conviction at seed reflected a view that the power constraint in AI was not a temporary problem.

The Series A followed in April 2025, a $70 million round led by Felicis and structured as a mix of $35 million in equity and $35 million in debt provided by HSBC Innovation Banking. Additional investors included Starwood Capital, Thrive Capital, 8090 Industries, MCJ Collective, MVP Ventures, GOAT VC, and StepStone Group, with Atomic and a16z both returning. Six months later, Exowatt raised a further $50 million extension led by MVP Ventures and 8090 Industries, bringing total capital raised to $140 million in under two years.

The investor base is worth reading carefully. Starwood Capital is one of the largest real assets managers in the world. Thrive Capital, founded by Josh Kushner, has a track record of backing category-defining companies early. HSBC providing debt at the Series A stage signals that institutional lenders, not just venture capitalists, are taking Exowatt’s revenue prospects seriously. And the fact that both Atomic & a16z came back in subsequent rounds is a solid nod for insider confidence.

What binds all of it together is the founding relationship between Happi & Abraham at Atomic. The venture studio model gave Exowatt something most hardware startups do not have in the early days: time and resources to iterate without the immediate pressure of external market expectations. Happi has described going through more than 50 different design configurations before landing on the P3. That iteration cycle, funded and sheltered inside a studio environment, is what produced a product that was ready for commercial deployment at the moment the market needed it most. The timing was not luck. It was the result of starting early, staying disciplined, and building in a city, Miami, that most of the industry was not paying attention to.

→ Listen on X, Spotify, YouTube, Apple

Subscribe now

NEW: Max Levchin, Affirm

Tue, 14 Apr 2026 01:16:10 GMT

Strikes & Gutters, Ups & Downs

Max Levchin, Fmr Co-Founder & CTO of PayPal, now Co-Founder & CEO of Affirm, joins Sourcery for a wide-ranging conversation on building companies, surviving failure, & navigating one of the most important economic shifts in decades, with a steady stream of The Big Lebowski references throughout.

→ Listen on X, Spotify, YouTube, Apple

Levchin is a legendary serial entrepreneur, technologist, & investor, best known for helping build PayPal (as CTO), Affirm, & backing (~100+?) startups including Brex.

We start inside Affirm HQ with a Max Masterclass in the Art of Espressos before going deep into lore of Affirm, Levchin’s career, & an AI macroeconomic outlook.

We cover:

Lessons from PayPal that still apply today
Why the team is the single biggest determinant of success
How Affirm onboarded 800K Shopify merchants in a week
Why you can’t perfectly time an IPO
Why now is the best time ever to be a CS CEO
How AI is collapsing the cost & speed of building software

One of Levchin’s key points: As intelligence becomes more accessible, weaker or “fine print” companies will get exposed faster. The baseline for what it means to build, operate, & compete is rising.. along with the effective IQ of the system.

This conversation was recorded March 30, 2026.

Subscribe now

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

(00:00) Max Levchin, Co-Founder & CEO at Affirm
(01:35) Inside Affirm's office espresso bar
(14:56) Meeting Alfred Lin & Tony Hsieh over poker
(21:14) Onboarding 800,000 Shopify merchants in one week
(22:59) Big Lebowski quotes hidden in every shareholder letter
(32:11) The PayPal lesson that built Affirm
(35:25) Being a technical CEO & why now is the best time to be a CS CEO
(42:10) Should engineers still learn to code?
(44:48) Side quests with AI
(46:59) Companies AI will destroy
(49:46) How AI has changed engineering at Affirm
(50:54) Agentic commerce & DoorDash
(52:28) Devolution of Credit
(55:06) The biggest misconceptions about buy now, pay later
(57:42) Being a public company CEO
(01:07:01) Advice for private companies
(01:11:09) Creating his own economy
(01:14:29) Can AI help solve the $39T debt problem?
(01:17:46) Will average IQ rise or fall?

Brought to you by:

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Merge—The leading provider of customer-facing integrations and agentic tools for frontier LLMs, Fortune 500 organizations, and B2B SaaS companies. Visit: https://merge.dev

“Fine Print” as a Business Model Is About to Be Destroyed..

Affirm CEO Max Levchin on the devolution of consumer credit, why every digital economy rediscovers capitalism, and what happens when AI gives ordinary borrowers the same leverage as Wall Street lawyers.

Max Levchin takes pride in crafting his own espressos (about five or six shots a day), timed & weighed to the gram. He quotes The Big Lebowski in shareholder letters & on-demand. He recently built an iOS app on weekends using Claude Code because he got annoyed at his home audio remote. Sets PRs on Strava on the daily. Co-founded PayPal. And he’s been CEO of Affirm for over a decade..

In a wide-ranging conversation at Affirm HQ, he covered the history of consumer credit, what AI does to opaque financial products, why every virtual economy converges on capitalism, and what it actually means to be a technical CEO right now.

How Lending Became Extraction

Credit, Levchin argues, started as a borrower-aligned product. Farmers needed capital between planting and harvest. Lenders provided it. The Code of Hammurabi included explicit prohibitions against overcharging borrowers. Early religious law in multiple traditions restricted or banned interest entirely. The original instinct of lending was protective.

The modern version reversed that.

~2000 BC: First documented borrower protections in Code of Hammurabi
100%: Effective gross margin on late fee revenue for major banks
50%: Share of Americans who carry revolving credit card balances
$48B+: Affirm loan volume projected this fiscal year

“At some point someone said, wait a second — that’s a hundred percent gross margin product. When I charge you a late fee, you just pay me more money. There’s no cost other than like sending you a nasty gram. Of course I’ll charge you a late fee. So it’s better if you’re late a lot, because then I’ll make more money and it costs me nothing to create that revenue line.”

Late fees became a near-costless revenue line. The industry optimized around them. The mechanism got buried in fine print. Affirm was built to counter that directly: no compounding interest, no late fees, no deferred rates. The total cost of a loan is shown upfront before the borrower accepts.

“We don’t compound interest specifically to make sure that we can pre-price every loan and say: you’re borrowing $500 and your total interest charges will be $25 or $0 if a merchant is paying your interest for you.”

Early on, consumers didn’t believe it. They’d been conditioned to look for the asterisk. That’s shifted. Affirm employees now regularly get stopped by customers who recognize the logo and want to talk about the product.

The PayPal Lesson That Stuck

Levchin met Alfred Lin & Tony Hsieh before any of them were known for anything. He was a college student, parts of a failed startup had been acquired by Link Exchange, and he flew to San Francisco to help with the integration. He showed up expecting a meeting. It was a poker game.

“I just fed myself and watched Alfred and Tony and a bunch of other people play poker. That was my introduction to the Bay Area.”

His early startups all failed. Rather than walk away, he kept going, less because of any grand vision, & more because starting companies was simply what he was going to do.

“I took exactly zero hours or minutes contemplating is this the right thing for me to do. I was like, oh, like I’m definitely gonna do this again.”

That persistence led to PayPal. As CTO, Levchin was at the center of what became the single densest concentration of founder and investor talent ever to come out of one company —Peter Thiel, Elon Musk, Keith Rabois, David Sacks, Roelof Botha, Reid Hoffman, and others, went on to found and/or lead the most consequential companies of the last two decades.

“The most important lesson is always the team.
It is the alpha & omega, success or failure of a company is the team.”

With the caveat:

“Just having a bunch of brilliant people is not actually enough. You need to organize them, give them a mission. If you don’t have the people, you’re definitely gonna fail. But even if you do have the people and you don’t harness their unique skills and correctly combine them, you might fail anyway.”

He puts the US startup failure rate at 85%. Most fail for people and alignment reasons, not technical ones.

To hammer down the point: The PayPal mafia isn't famous for the exit. It's famous because the talent never stopped compounding.. from founding, building, and backing the next generation of great companies, literally creating trillions in value across decades.

That's the real measure of a successful team. Whether the culture you build keeps generating talent, ideas, and companies long after people have moved on.

Why Every Digital Economy Rediscovers Capitalism

Levchin grew up in Soviet Ukraine. He watched a command economy fail firsthand. His view on capitalism isn’t ideological, it’s observational. Markets work. He’s spent his career inside ones that do.

When asked how he'd construct a new economy from scratch, like the closed digital systems a la Roblox, Second Life, and others.. a common theme emerged. Every one of them eventually arrived at the same structural solutions. They didn't copy the US economy. They independently arrived at it.

"It may be lack of imagination, but if you look at the success arcs of these systems over time, they resemble the US economy more and more. Sort of like basically start out with a fixed money supply and like, well, we'll figure out how to inject more money later. And then suddenly more people come in and you need more liquidity. And so they sort of start pegging their currency to the dollar or floating it — it becomes more and more like the real world."

His answer on constructing a new economy: he wouldn’t. Capitalism works. We’ve had thousands of years to evolve it and the US got it right.

On the $39 trillion national debt — so long as the US produces enough value to compensate those that hold it, the position holds. The follow-up question was whether AI-driven productivity could help cover the difference:

"It's hard to tell. It's a really good question actually, & the question is, where does value accrete in this new age? On the one hand, it's very tempting to say, well, for the country of our resources & our access to physical resources & capability that we have to build and innovate and sort of create things out of atoms, having this incredible boost of productivity expressed in bits is just gonna be incredibly powerful & leapfrog anything and everyone that is remotely buying our supremacy. You could also make the argument that all these countries where there's precious little by way of resources are now boosted tremendously by advances in AI.
I tend to believe in the former, just because I'm being an American by choice. Better be a patriot, otherwise what exactly were you thinking? I tend to think that we'll probably end up with just an amplification of our advantage. That said, we're not without rivals on the world stage and we'll have to compete. But I'm a big fan of competition."

The End of the Fine Print Business Model

Most fintech commentary on AI focuses on underwriting, fraud detection, and operational efficiency. Levchin’s core argument is different: AI is about to kill an entire category of business model.. specifically, any model that depends on consumers not fully understanding what they signed up for.

“No AI is ever going to be like, ‘Oh, you’ve given me a task to find you a good loan. I found one that’s going to take advantage of you. Hooray.’ Absolutely not. I’m not going to let that happen to you. My job as a robot is to take good care of you.”

AI agents acting on behalf of consumers will read terms & conditions that consumers currently don’t. They’ll compare rates across products without fatigue. They’ll catch deferred interest clauses, variable rate triggers, & penalty structures buried in the fine print.

The information gap that consumer finance has operated on for a century — institutions with lawyers on one side, borrowers without time or expertise on the other — closes when every borrower has an AI reading on their behalf.

Levchin firmly believes this will lift the average IQ to a higher standard.

“The average IQ is still a hundred. I think the average IQ with AI in your ear at all times is about to go up to 150 — which is north of the genius definition.”

For Affirm, this is a structural advantage. A company built entirely on transparency has nothing to hide from an AI agent doing comparison shopping. Companies whose revenue depends on consumers not understanding their products are in a different position.

Why CS Still Matters in an LLM-First World

Levchin has been writing code since he was eleven or twelve. He planned to get a doctorate in computer science before startups interrupted that. He still reads AI research papers nightly, not as homework, but because the gap between a paper publishing and someone shipping an implementation has essentially closed. Missing a week of papers means missing something that will matter.

“It is the moment in time to be a CEO with a computer science degree. You can now go from I have an inkling in my head to I’m going to build a prototype to I’m going to ship something that’s production ready — in record short time.
If you do know what you’re doing as somebody with a sense for how to build software, you don’t need to vibe code yourself to a prototype and hand it off. You can actually build something amazing and just ship it.”

On whether deep CS knowledge (like idk the hot topic of coding?) still matters when LLMs write code on demand.. his answer is yes, but for a specific reason.

Software engineering is a science, an art form, and a craft simultaneously. The science is teachable. The art and craft require taste. Taste can’t be fully delegated to a model. You still need to know what elegant code looks like, recognize when you’re not getting it, and be able to steer the model toward something better.

Without that foundation, you accept whatever output comes back. Which might work. But probably isn’t great.

Long-Term Thinking in a Quarterly World

Affirm was ready to go public in under ninety days from the decision to file. The SEC asked them to wait too many companies were rushing to market at the same time. They went public in early 2021.

"In December of 2020, the SEC told us, please, don't come out this year. We're so overwhelmed by all these companies trying to go public. Could you please hang on and do it next year."

While many are overwhelmed by the process of going public, he says it’s actually more manageable than people make it out to be (& maybe they should take the advice Reddit CEO Steve Huffman got: ‘don’t be a little bitch about it’). What’s harder is the discipline not to let quarterly reporting become the company’s primary focus. He’s seen CEOs lock themselves in conference rooms for a week before earnings, obsessing over a period that ended six weeks prior. The quarter is already over. The only useful question is what comes next.

He referenced the Graham/Buffett line: “in the short term the market is a voting machine, in the long term a weighing machine.” Levchin said it’s what he reminds himself of whenever price action moves in either direction. Up or down, the answer is the same: keep building.

Strikes & gutters, ups & downs.

He has no interest in shareholders focused on short-term price action. Affirm is estimating north of $48 billion in loans annually and still represents a fraction of a percent of total US commerce. The company, by his math, has decades before the final grade is in. The IPO is just another data point in the journey.

The Dude Abides.

A running Easter egg in every letter

For twenty-plus quarters as a public company, Affirm’s shareholder letters have contained a Big Lebowski reference. Not once. Every single one. Levchin keeps the corrected Lebowski script open on his desktop while drafting them so he can search for quotes he hasn’t used yet.

Until recently, he wasn’t getting much feedback on them.. a fun hunt carefully curated for each letter. Then, about three quarters ago, an analyst published a note with the line: “Q2 fiscal 25 — a quarter that really tied the room together.”

“Chef’s Kiss”

He’s seen the film enough times that quotes surface unprompted mid-conversation. When asked about his media diet he said, unprompted: “The occasional acid flashback.” “that’s a quote from Lebowski. I have not in fact had any acid flashbacks that I can think of.”

He recently watched it with his fourteen-year-old daughter for the first time. His sixteen-year-old son was inducted a couple years prior. The Dude is in the DNA.

On the cast:

“Philip Seymour Hoffman, very young Philip Hoffman, so he’s kind of funny and spastic. Jeff Bridges is not so bad. Very young John Turturro playing a deranged bowler named Jesus.”

On why it holds up:

“It’s one of these cult movies where you watch it and you’re like, this is amazing, and you watch it again, like no, it’s better than amazing.
You watch it again, it keeps on getting better. Keep on finding these gems.”

His favorite kind of Lebowski moment is the cultural deep cut that has almost nothing to do with the plot. He described one specific scene — the Dude explaining something to Walter, Walter riffing on Lenin, and Donny popping in:

“Donny pops in and says ‘I am the Walrus.’
He keeps on popping in ‘cause he thinks Lenin being referred to is actually John Lennon.”

Then, mid-tour, standing near the office game station, entirely unprompted:

“Also, dude, having an amphibious rodent as a domestic — that ain’t legal either. What are you a fucking punk ranger now?”

And later:

“You said it, man. Nobody fucks with the Jesus.”

On whether any film made today will earn the same status:

"The real litmus test of a great beautifully written movie or well written anything is it survived a test of time.
So I know, but I'll take another ten years. I'm sure there's some amazing scripts that are going to be discovered as the great cult classic of 2026 that are in theaters right now, but we won't know until 2036."

The Dude abides.

→ Listen on X, Spotify, YouTube, Apple

Subscribe now

Sourcery

BREAKING: Inside Impulse Space w/ Tom Mueller (FULL TOUR)

SpaceX’s #1 Employee Built the Merlin Engine, Now Building Impulse Space

→ Listen on X, Spotify, YouTube, Apple

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

Brought to you by:

Inside Impulse Space with Tom Mueller: Engines, Spacecraft, & the Case for Moving Things in Orbit

→ Listen on X, Spotify, YouTube, Apple

From Merlin to Impulse

Mira: The Precision Maneuvering Spacecraft

Helios: The Vehicle That Gets You There

The Deneb Engine

How Impulse Builds

Operating in Orbit

SpaceX Lessons: Starlink, Iteration, & Elon

The Long View: Nuclear, Orbital Data Centers, & the Moon

→ Listen on X, Spotify, YouTube, Apple

Want More Impulse??

Check out our full interview w/ President & COO Eric Romo

→ Listen on X, Spotify, YouTube, Apple

ICYMI: Out of 50K+ Global Startups, 5 Won The $1M Pitch

Brought to you by:

ICYMI: Inside Deel’s Pitch Grand Finale

Five Startups, +50,000 Applications, & a $1M Bet on Borderless Founders

Why This Competition Comes From Deel, Specifically

The Tournament That Funnels 50,000 to 5

The Judges and the Money Behind Them

The Five Finalists & AI

The Bigger Reveal: Deel Ventures

What It Says About Seed-Stage Funding in 2026

Want more from Deel? Watch our interview w/ CEO Alex Bouaziz:

→ Listen on X, Spotify, YouTube, Apple

BREAKING: Impulse Space Raises $500M Series D

Why Impulse Raised $500M to ‘Move Things in Space’

→ Listen on X, Spotify, YouTube, Apple

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

Brought to you by:

Impulse Space Series D: An Investor Briefing

→ Listen on X, Spotify, YouTube, Apple

The Founding Story

The Round

The Investors

Market Growth: $17 Billion → $71 Billion

Product: Mira

Product: Helios

Product: Caravan

The Propulsion Family

Use of Proceeds

→ Listen on X, Spotify, YouTube, Apple

ICYMI: "We Don't Trust Agents" - Here's How to Control Them

‘Agent Handler’ is now available

Control AI Access, Tokenmaxxed $$$ Bills, & Stop Mass Data Leaks

→ Listen on X, Spotify, YouTube, Apple

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

Brought to you by:

Inside Merge: How a Data-Sync Startup Became Infrastructure for Enterprise AI

→ Listen on X, Spotify, YouTube, Apple

The Bet That Reshaped the Company

The Three-Product Suite

Agent Handler & the Employee AI Problem

A Cautious View of Agents

How AI Companies Actually Buy

The Cost Problem & the Case for Routing

The SaaSpocalypse & Road Ahead

→ Listen on X, Spotify, YouTube, Apple

BREAKING: Inside Lead Bank - $56M Investment Now Worth $1.5 BILLION

Jack Ma, Masa, Jack Dorsey, Jackie Reses - The Story Behind Lead Bank

→ Listen on X, Spotify, YouTube, Apple

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

Brought to you by:

The Jackie Reses Playbook: Inside the Career of Fintech’s Most Iconic Builder

The Improbable Career Arc: From Atlantic City to the Top of Silicon Valley

→ Listen on X, Spotify, YouTube, Apple

Inside the Alibaba Boardroom: Jack Ma, Joe Tsai, Masayoshi Son, + Jackie

Yahoo & the HR Job That Wasn’t an HR Job

Square, Jack Dorsey, & the Hack Week That Built a Bank

Lead Bank: Buying a 100 Year Old Charter to Rebuild Fintech Infrastructure

The Team Behind Lead Bank

Architecture & AI: The Five APIs Powering Modern Fintech

Crypto, De Banking, & Why Jackie Calls BS on the Narrative