Dylan Patel, founder and chief executive of SemiAnalysis, said on Fox Business on 12 November that Nvidia’s leadership change “bodes really well” for Anthropic’s IPO, because weaker open-source models leave Anthropic “much more likely to get a complete monopoly or a duopoly on this ecosystem”. Anthropic’s own filing shows the slowdown Patel pointed to: its quarterly revenue rose $5.4 billion in the third quarter, against $6.8 billion in the second, a gain about 21% smaller, by our arithmetic.
The same filing has a later number. Anthropic’s run-rate revenue rose $10 billion in that single month, to about $81 billion. And it names four chip families it trains and serves Claude on, which is the base of Patel’s second argument, about Google’s TPUs.
Business Pill · Run-Rate Revenue Explained
An evergreen Business Engineer explainer on run-rate revenue, what an AI company’s annualised number means and how it differs from reported revenue. It teaches the concept only and does not discuss Anthropic’s filing or this segment.
The key insight: As we read the filing against the clip, the slowdown Patel described shows up in Anthropic’s own table: the quarterly revenue gain fell to $5,401 million in the third quarter from $6,824 million in the second, 20.9% smaller, by our arithmetic, before run-rate revenue rose $10 billion in October alone.
The host, Charles Payne, asked “what does this mean maybe next week for the Anthropic IPO?” Patel said the recent turmoil in tech stocks was, in Patel’s opinion, “driven off of Anthropic’s revenue slowing down”.
In the second quarter, Patel said, “Anthropic plus OpenAI added about $33 billion of revenue, ARR”. In the third it was less, and “that was primarily off the backs of open source gaining so much steam”, Patel said. With Nvidia’s ability to fund open models in question, Patel said “open source looks a lot weaker”.
A clip of the answer, posted on X by Fireside Alpha on 13 November, had about 16,000 views when we read it. This piece covers the IPO and compute passages; our earlier piece on Patel’s case for renting chips to China covers a separate SemiAnalysis episode.
Anthropic’s quarter-on-quarter revenue gain: $658 million in the fourth quarter of 2025, $2,691 million in the first quarter of 2026, $6,824 million in the second and $5,401 million in the third, 20.9% smaller than the second. Source: Anthropic’s amended S-1 (Amendment No. 2, 10 November 2026), quarterly revenue table; the gains and the percentage are our arithmetic.
What Was Posted on X
SemiAnalysis’s Dylan Patel says Jensen’s exit bodes really well for Anthropic’s IPO, because a weaker open source ecosystem leaves Anthropic closer to a monopoly or duopoly
“I think it’s no surprise that the big turmoil in the market in tech, in my opinion, was just driven off of Anthropic’s revenue slowing down. In Q2, Anthropic plus OpenAI added about $33 billion of revenue ARR. And in Q3, it was less, and that was primarily off the backs of open source gaining so much steam.”
“Now open source looks a lot weaker with Nvidia down and out, at least in some sense, their ability to fund the open source. And if that’s the case, Anthropic looks like they’re much more likely to get a complete monopoly or a duopoly on this ecosystem. So I think this bodes really well for Anthropic’s IPO next week.”
“And furthermore, they’re the one company in the world that knows how to use TPUs. They know how to use Amazon’s Trainium. They know how to use AMD GPUs, and they’re building their own chip. They’re more diverse in what chips they use than any other company in the world. Their fights with Nvidia have been brewing for a long time. They figured out how to exist and use compute without Nvidia. Of course, they still use a lot of Nvidia, but two thirds of their compute is not Nvidia. And so they’re well positioned to take advantage of a weaker Nvidia and a weaker open source.”
Fireside Alpha posted this clip at 04:06 UTC on 13 November, from Dylan Patel’s appearance on Fox Business’s Making Money the day before. It carries Patel’s remarks on Anthropic’s IPO, open-source models and chip mix; this piece quotes Patel only where our transcript of the broadcast matches the post’s text, then sets the figures against Anthropic’s amended S-1.
What Anthropic’s Filing Shows on Revenue
Anthropic’s amended S-1, filed on 10 November, lists quarterly revenue of $4,731 million for the first quarter of 2026, $11,555 million for the second and $16,956 million for the third. The quarter-on-quarter gain was $6,824 million in the second quarter and $5,401 million in the third, by our arithmetic.
That is $1,423 million less, or 20.9% smaller. Revenue still grew 46.7% in the third quarter, after 144.2% in the second, by our arithmetic. Multiplied by four, the two gains come to about $27.3 billion and $21.6 billion a year.
The filing then gives a monthly figure. Its run-rate revenue “was approximately $81 billion as of October 31, 2026, up $10 billion from September 30, 2026”. That puts 30 September at about $71 billion, by our arithmetic. The October gain alone equals about 46% of the third quarter’s annualised revenue gain, though the two measures are built differently.
The filing itself warns about both measures. It says “we expect our Run-Rate Revenue growth rate to decline over time”, and that run-rate can move with “reduced usage we have typically experienced around holiday and vacation periods in the summer and winter”.
Where OpenAI Fits
OpenAI is private and does not file results. Bloomberg reported that its revenue rose to $10 billion in the third quarter from $6.7 billion in the second, as we reported, a gain of $3.3 billion. On those figures, Anthropic’s third-quarter gain of $5.4 billion was about 1.6 times OpenAI’s, by our arithmetic.
Patel’s $33 billion is a run-rate figure for the two companies together. The amended filing we read gives Anthropic’s run-rate for 31 October and, through its comparison, 30 September; we found no 30 June run-rate figure in it. So the $33 billion cannot be rebuilt from the filing, and we do not try.
The Compute Claim: Four Chip Families
Patel said Anthropic is “the one company in the world that knows how to use TPUs”. “They know how to use Amazon’s Trainium. They know how to use AMD GPUs, and they’re building their own chip”, Patel said. “Of course, they still use a lot of Nvidia”, Patel added, putting the share of Anthropic’s compute that is not Nvidia at about two-thirds.
The filing says: “We train and serve Claude natively across four major chip families (AWS Trainium, Google TPUs, NVIDIA GPUs, and AMD GPUs)”. We found no split of compute by chip vendor in it, so the two-thirds figure is Patel’s estimate.
The commitments show the scale. Anthropic agreed to spend at least about $121.3 billion with Google between April 2026 and July 2033 under a hosting agreement, and estimates a further $125 billion to $150 billion for additional TPU compute delivered between 2028 and 2031. It committed at least $110.0 billion to Amazon and $47.2 billion to $49.2 billion to Microsoft. All commitments total about $900.9 billion, against $131.8 billion of cash and short-term investments on 30 September, as we reported. Compute was also the theme when Amodei spoke with Jamie Dimon on Friday.
Nvidia is also a buyer in the deal. At the $74 midpoint, NVIDIA would buy about $10.0 billion of Class N stock in the concurrent private placement, the same as Google, out of $26.0 billion from seven companies. Our piece on the $1.25 trillion to $1.4 trillion target covers the price range.
Anthropic’s compute spend by provider, from its amended S-1 of 10 November 2026: at least about $121.3 billion with Google from April 2026 to July 2033, an estimated $125 billion to $150 billion for additional TPU compute delivered between 2028 and 2031, at least $110.0 billion with Amazon from May 2026 to April 2036, and $47.2 billion to $49.2 billion with Microsoft from November 2026 to June 2033. Bars show the low end of each range; all commitments total about $900.9 billion.
Patel on Google’s TPUs
Earlier in the same segment, Patel described SemiAnalysis’s own testing: “we run benchmarks on them on open source models every night. They’re just as good as Nvidia”. Without Nvidia’s founder making supply deals, Patel asked, “does that now mean Google can just take a ton of share?” These are Patel’s views.
The Structural Read
As we read Patel’s argument, it ties the IPO to who funds open models rather than to Anthropic’s own numbers. The filing lists open-weight and open-source models as competition for cost-sensitive work, but in the passages we read it does not say why the third quarter slowed.
As we read the compute passage, the filing supports the breadth Patel described: four chip families and at least $121.3 billion committed to Google. The two-thirds non-Nvidia share stays Patel’s estimate, and NVIDIA is still set to buy about $10.0 billion of the stock.
Anthropic, amended S-1 (Amendment No. 2)
“We train and serve Claude natively across four major chip families (AWS Trainium, Google TPUs, NVIDIA GPUs, and AMD GPUs)”
Three Implications
FOR IPO BUYERS The quarterly table shows the third-quarter revenue gain fell to $5.4 billion from $6.8 billion; the $10 billion October rise in run-rate revenue is the latest figure the filing gives.
FOR CHIP SUPPLIERS Anthropic names four chip families and at least $121.3 billion of Google hosting spend to July 2033; NVIDIA would buy about $10.0 billion of its stock at the $74 midpoint.
FOR OPEN-MODEL PROVIDERS Anthropic’s filing names open-weight and open-source models as competitors for workloads that prioritise cost; Patel ties the third-quarter slowdown to their growth.
The framework sets out the pattern: “Close the SCARCE layer (keep proprietary) + Open the ABUNDANT layer (commoditize)”.
As we read it, Patel’s argument is about who pays for the open layer: if Nvidia funds less of it, closed model makers such as Anthropic keep more of the scarce one. The framework was published in March 2026, before the filing.
What Is Not Established
The filing does not test Patel’s explanation. In the passages we read, Anthropic does not attribute the third-quarter change to open-source models; it does say open-weight and open-source models “compete for a growing range of enterprise workloads that prioritize cost efficiency over intelligence”.
The $33 billion combined figure and the two-thirds compute share are Patel’s, spoken on air, and are not in the filing. OpenAI’s figures are reported, not filed. The IPO has not priced. Quotes are from our transcript of the broadcast, checked against the clip’s text.
This Week In The AI Supercycle
The Great Convergence of Capital, Power, and Geopolitics
This week’s issue: for two years the AI build-out was priced as a supply story, and last week as a credit story. This week it was priced as a permission story, and the answer arrived as a cost, not a ban.
Patel argues that a weaker open-source camp helps Anthropic’s IPO. The filing shows the smaller third-quarter gain Patel described, $5.4 billion after $6.8 billion, and then a $10 billion run-rate rise in October. It lists the four chip families, with at least $121.3 billion committed to Google. The $33 billion and the two-thirds share remain Patel’s estimates.
95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. We traced the clip Fireside Alpha posted on X on 13 November to its source, the Fox Business segment of Making Money with Charles Payne published on 12 November 2026, and transcribed the broadcast audio ourselves. We quote Dylan Patel only where our transcript and the text of the clip post carry the same words. We read Anthropic’s amended S-1 (Amendment No. 2, filed 10 November 2026) on SEC EDGAR, including its quarterly revenue table, run-rate disclosure and compute commitments.
OpenAI’s quarterly figures are as Bloomberg reported them; OpenAI does not file results. The quarterly gains, growth rates, annualised figures and ratios are our arithmetic. We leave out the segment’s personal material, succession talk and other claims. We did not contact SemiAnalysis, Fox Business, Anthropic or OpenAI for this piece.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
