What to expect..
I sat down with Dylan Patel, Founder & CEO of SemiAnalysis, to discuss AI markets, and how he's building a research and investment empire with the acquisition of Citrini Research and the launch of a $500 million venture fund.
→ Listen on X, Spotify, YouTube, Apple
Dylan has built SemiAnalysis into one of the most influential independent research firms covering semiconductors, AI infrastructure, data centers, and global supply chains. Today, he oversees 115 people across SemiAnalysis & Citrini, with businesses spanning proprietary data, technical research, consulting, & now venture capital.
We discuss the competition between OpenAI, SpaceXSI, Anthropic, Google, and Meta, the shift from foundation models toward applications, and how changing product adoption, margins, and valuations are shaping the AI market.
Dylan also breaks down the economics of AI infrastructure, the risks facing data center development, credit markets, AI valuations, and where he sees opportunities emerging across the technology stack.
We then get into the business behind SemiAnalysis: why he acquired Citrini Research, how he's scaling a global research organization, and why he raised a $500M solo-GP venture fund to invest at the seed and Series A stages across semiconductors, robotics, optics, networking, manufacturing, and infrastructure software.
𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒
(00:00) Dylan Patel, Founder & CEO at SemiAnalysis
(00:37) Inside the Hogwarts library
(02:17) Are data centers in space the answer?
(06:42) Who’s really against data centers?
(08:47) Elon's data centers vs. 3 In-N-Outs
(12:04) Who's winning: Anthropic, OpenAI or Google?
(13:46) Why Meta's Muse beats OpenAI on product
(15:55) Where SemiAnalysis spends $15M on AI
(20:38) Is there real ROI in AI startups?
(22:30) Inside Dylan's $500M fund
(25:22) Why Dylan won't invest in apps
(28:43) Most pitches are boring and wrong
(30:30) Buying Citrini, the AI bears
(33:20) Sending an analyst to the Strait of Hormuz
(34:03) Why Gartner is a racket
(41:07) The hardest lessons of running SemiAnalysis
(43:29) Knowing just enough Chinese
(46:32) Why hardware is about to explode
(48:03) Why Dylan sees better in the dark
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 OpenAI, Anthropic, Vercel, Granola, Deepgram, & Sourcery.. teams that move fast AF. visit → brex.com/sourcery
Harvey—The AI platform built for legal work, is helping the best firms and companies close deals faster. More than 200,000 lawyers, 80% of the Am Law 100, and over 20% of the Fortune 500 build with Harvey. Visit: Harvey.ai
Zone—develops next-generation data center campuses, partnering with AI companies, site developers and technology leaders to bring compute online faster and at scale. Visit: zonefrontier.com
Mercor—The human intelligence behind frontier AI. Frontier labs and leading enterprises use Mercor to turn the knowledge of 5M+ experts into the data, evals and benchmarks that make AI better at real work. Visit: http://mercor.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 is the world's first agentic brokerage. You can invest across stocks, options, bonds, crypto, event contracts, & more, then automate anything with AI Agents. Seriously, just tell an Agent what you want—from "Set up a buy-the-close, sell-the-open strategy for me" to "Automatically sweep any cash over $10k from my bank to my brokerage account." Agents can monitor markets, run strategies, & manage your cash 24/7. Get started: public.com/sourcery
Dylan Patel on Building SemiAnalysis, the Citrini Deal, & a $500M Solo GP Fund
Dylan Patel founded SemiAnalysis in 2020 as a semiconductor research newsletter. The firm now sells data, consulting and research to hyperscalers, AI labs, chip companies and infrastructure funds. In September 2026 SemiAnalysis acquired Citrini Research. Patel has also launched SemiAnalysis Capital, a $500M early stage fund that he runs as a solo GP.
SemiAnalysis and Citrini have 115 people across 15 countries, with offices in New York, Chicago, San Francisco, Oregon, Singapore and Korea.
→ Listen on X, Spotify, YouTube, Apple
SemiAnalysis Tracks Every Data Center, Chip, & Cable Shipment
Most SemiAnalysis revenue comes from 2 lines, data sets sold to clients and one-on-one consulting projects. The consulting work is kept separate from the published research. The firm tracks every data center in the world, along with build timelines, chip shipments, memory pricing, and the tools going into fabs.
The research is bottom up. SemiAnalysis tests chips, clusters and models in-house, and it recently launched GDPX, a benchmark covering cyber, chip design, hardware design, materials science, bio, quant and RSI.
“We’re tracking every bit of supply chain, how many cables are shipping, how many servers are shipping, how many GPUs, how many tools are shipping for the fabs.”
The staff includes people from ASML, Applied Materials, Lam Research, and TSMC and Intel R&D. Others come from infrastructure roles at Google, Microsoft and Amazon, or have trained models, built compilers, built data centers and built power plants. Patel says the firm knows more about data centers and behind-the-meter power than anyone else.
Permits Have Pushed Gigawatts From 2027 Into 2028
Permitting is now the biggest risk to the data center buildout. SemiAnalysis has moved several gigawatts of capacity from 2027 into 2028, and more is slipping out of 2028. Construction capability, labor and equipment are available for these projects. The permits are not.
“We’ve pushed back multiple gigawatts out of 2027 into 2028.”
In Texas, Patel sees Governor-elect Gina Hinojosa’s electricity price and water conditions as solvable. He sees her union labor requirement as the open question in a state with little union labor. He notes the governor controls water, wastewater, stormwater and air permits. The air permit alone can stop a data center that isn’t on the grid and uses little water.
Patel puts gigawatt-scale capacity in space at around 2030 at the earliest. He doesn’t expect it to add meaningful capacity in the next 3 to 4 years. He takes the over on Elon Musk’s 2028 targets.
49% of People Near a 100 Megawatt Data Center Don’t Know It Exists
SemiAnalysis ran a poll with David Shor across every county with a 100 megawatt or larger data center, built, under construction, or planned within 3 years.
“49% of people in counties that they have 100 megawatt data center did not even know there is a data center in their county.”
Only 40% of people in counties with a 100 megawatt site planned knew it was coming. Data centers did not rank among the top election issues in exit polls, which led with inflation and the war in Iran. Most of the places that have banned data centers or passed moratoriums had little or no capacity going up.
Patel says modern data centers use little water when built well. Counting embedded water across the supply chain on both sides, he says 3 In-N-Out locations use as much water as all of Elon Musk’s data center capacity. A New York Times bestseller overstated data center water use by about 1,000x and later revised the figure. He also points to a Louisiana county that paid each of its teachers about $20K from data center tax revenue. A SemiAnalysis post found that subsidizing power for everyone within a 30 mile radius costs operators very little.
45% of a $15M AI Budget Now Goes to Persistent Agents
SemiAnalysis spends about $15M a year on AI. For most of the year Claude Code took the largest share. Over the last few months, persistent agents went from zero to 45% of total spend. Most of that runs through Perplexity Computer, which uses Claude and OpenAI models behind its own harness.
Patel says Meta’s Muse and Instinct are better products than OpenAI’s Dots even though they run on weaker models. He points to OpenAI’s Guardian safety layer, which he says blocks tasks in Codex and Dots too often, and to the model and reasoning settings in OpenAI’s app.
“Instinct doesn’t have the best model. They’re primarily using open models. And yet OpenAI is terrible versus them.”
On stickiness, Patel expects consumer products like Muse and Instinct to keep users more easily than coding tools, since developers switch freely. Enterprise contracts lock customers in more. He has seen Anthropic offer enterprise customers discounts that depend on sending 70% or more of their AI spend to Anthropic.
Patel says public valuations move less than the headlines suggest. On the private side, Anthropic led in Q2, OpenAI led in Q3, and Anthropic is back ahead on Opus 5.5 and Fable 5.2. He expects most AI startups to fail. He cites Sierra and Harvey for fast revenue growth, & he flags pass-through margins as a risk for application companies.
We recently chatted with Apoorv Agrawal AI Partner at Altimeter where he laid out the “Four Seasons of AI” to make sense of the change up in who is ‘top dog’ at the moment with all the models racing against each other. It’s a worthwhile listen and shares the competitive dynamics, strengths, & weaknesses of each.
Why Patel Started a $500M Fund
Patel lists his goals as knowing everything, having influence, making money, having fun, and spending time with family. A research firm serves the first 2. Selling data and consulting covers the third. The fund adds a way to act on companies SemiAnalysis sees early.
“If you see something awesome super early, we started an early stage venture fund. I raised $500 million solo GP.”
The fund writes seed and Series A checks across infrastructure, from robotics to silicon, optics, manufacturing, networking, and software for hardware. It skips consumer applications. Patel says he doesn’t know how to scale an app to a billion users, so he has no edge there.
“I don’t know shit about applications, so why would I invest in applications?”
He says hardware needs both a strong founder and a strong idea. In application software, a strong founder can find the idea later. Check sizes are confidential, which Patel says keeps his options open. Deals come through his network and referrals from the 115 people across SemiAnalysis and Citrini.
How the Fund, Research, & Consulting Stay Separate
The fund is its own legal entity, walled off from SemiAnalysis and Citrini. Information gathered under consulting contracts stays covered by those contracts, and Patel says the firm spends heavily on compliance. Neither research firm will promote companies in the fund’s portfolio. Founders get help through introductions, hiring and research access.
“We’ll never use the platform I built to pump the companies I’ve invested in.”
Patel points to public calls against his own interests. He has a large investment in FluidStack, and SemiAnalysis has publicly said FluidStack’s data centers are delayed. The firm rated IREN’s cloud technology poorly in ClusterMax and still noted the value of its powered land. It has also published criticism of product teams at OpenAI and Anthropic, both large customers.
“I have a large investment in FluidStack, and yet we’ve still said their data centers are delayed publicly.”
SemiAnalysis has never run ads or sold an opinion. Its biggest customers are hyperscalers, major semiconductor companies and AI labs, and its research criticizes each of them at times. Some clients have tried to pull contracts over published views, and their own teams kept the data subscription. Patel contrasts this with Gartner, citing a Magic Quadrant that ranked IBM’s coding agent above Cursor or Claude Code.
Citrini Stays Independent & Can Stay Bearish
SemiAnalysis announced the Citrini acquisition on September 11, 2026. Citrini remains a separate company. James van Geelen stays as CEO, and director of research Kyle has the most say over what gets published. Patel’s plan is to make Citrini operationally stronger and leave its research voice alone.
“Dude, do whatever you want. I’m not here to change your voice. I’m here to help you run this as a better company.”
The 2 firms disagree on AI. Citrini was very bullish on AI and turned bearish a few months ago. Patel remains bullish on AI over the long term. Its February 2026 report moved markets and was presented as a thought exercise, and Patel says some of its points have since come true.
Patel describes the research methods as different. SemiAnalysis works technically and numerically, chip by chip and data center by data center. Citrini goes into the field. It developed a scorpion antivenom with AI to test AI drug design, and it sent someone to the Strait of Hormuz to assess oil shipping.
Running 115 People Across 15 Countries
Patel says the hardest parts of running SemiAnalysis are the decisions. They include customer and supplier negotiations, disputes between team members, hiring, firing, and a lawsuit from a former employee. He uses that experience to separate himself from investors who have only made investment decisions.
“A lot of investors have never made tough decisions.”
He describes much of company building as learning by doing, with few playbooks to follow. He knows of no model for the combination SemiAnalysis is building across research, consulting, media and venture.
Chip Testing, GDPX, & Faster Hardware Development
Over the next 12 months, SemiAnalysis is testing inference performance across Huawei chips, Amazon chips and chips from AI startups. Current models score poorly on most GDPX domains. Patel expects a sharp rise in the number of hardware companies as models improve.
“As the models get better at them, what does that unlock for society?”
He’s got his eyes on robotics, niche semiconductor work in packaging, networking & memory.
→ 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 Ad. Agents are an experimental AI tool. Its output may be inaccurate and is not investment advice. Public Holding’s, Inc. makes no guarantees about its accuracy or reliability—verify independently before use. Brokerage accounts & services provided by Open to the Public Investing, Inc. See additional disclosures at: https://public.com/disclosures/agenticdisclosures.

















