The Information reported on 3 November that Anthropic has told investors in recent meetings that its new 15-person chip design team is already using Claude for engineering and design work on its custom chips, citing four people familiar with the discussions. The meetings come as Anthropic heads toward an initial public offering that The Information says could value it around $1.5 trillion.
The 15 people are about 0.3% of the roughly 5,300 employees Anthropic reported at 30 September in its amended S-1, by our arithmetic. The bill such chips would address is larger: the filing puts compute and infrastructure expenses at $9.86 billion in the third quarter, up from $944 million in the first quarter of 2025.
Business Pill · Vertical Integration
An evergreen Business Engineer explainer on vertical integration: why every layer takes a cut. It teaches the general idea only and does not discuss Anthropic or its chips.
The key insight: As we read it, the new element is that Anthropic is pitching its own model as a tool for building the hardware it rents today: the S-1/A lists third-party chip design firms as a dependency for its custom accelerators, while The Information reports a 15-person in-house team already using Claude on its custom chips.
According to The Information, Anthropic’s engineers have found that Claude cuts down the time it takes to design chips compared to their estimates of traditional workflows at large chip designers such as Nvidia and Advanced Micro Devices. The outlet attributes that to some of the four people.
The comparison is an estimate by Anthropic’s own engineers as reported by The Information, not a measured benchmark. The free part of the article gives no time saving, no chip name, no foundry and no date for a first chip. The rest of the article is behind The Information’s paywall, and we have not read it.
Walter Bloomberg (@DeItaone) relayed the claim on X at 14:05 UTC, five minutes after publication, crediting The Information. Techmeme also carried the report, summarising it as “Sources: Anthropic tells investors that its 15-person chip design team is already using Claude for engineering and design work on its custom chips”.
Anthropic’s quarterly compute and infrastructure expenses and revenue, Q1 2025 to Q3 2026, as stated in its amended S-1 filed on 2 November 2026.
What Was Posted on X
ANTHROPIC ENGINEERS FOUND CLAUDE CUTS TIME TAKEN TO DESIGN CHIPS COMPARED TO THEIR ESTIMATES OF WORKFLOWS AT DESIGNERS SUCH AS NVIDIA, AMD- THE INFORMATION
Walter Bloomberg (@DeItaone) relayed The Information’s report at 14:05 UTC on 3 November. The time saving it describes is the Anthropic engineers’ own estimate against their view of workflows at Nvidia and AMD, as reported by The Information’s sources, not a measured benchmark.
Anthropic is now using Claude to help design its OWN AI chips.
According to a report from The Information, Anthropic is pitching this to IPO investors as a new strength of its models.
The company has a chip team of just 15 people, and that team says Claude is already speeding up its engineering and design work.
The chips would be used to train and run Claude, and Anthropic has not announced any plans to sell the design tools to anyone else.
This matters because compute is by far Anthropic’s BIGGEST cost and its amended IPO filing lists about $895 billion of commitments, and nearly all of it is for computing and data centers.
Broadcom is the largest name on that list with about $161 billion of equipment leases, followed by Google at about $121 billion and Amazon at $110 billion.
Right now Anthropic runs Claude on Google TPUs, Amazon Trainium, Nvidia GPUs and AMD GPUs, and it treats all of them as one shared pool of compute.
A custom chip would give Anthropic more control over that bill, and using its own AI to design it could make the chip cheaper and faster to build.
If a team of 15 people can do the work of a much bigger chip team with Claude’s help, that is a big signal for the entire semiconductor design industry.
Milk Road AI, citing The Information, says Anthropic is pitching IPO investors on Claude speeding up the engineering and design work of its 15-person chip team, with the chips meant to train and run Claude and no announced plan to sell the design tools. The post ties this to the roughly $895 billion of commitments it says Anthropic’s amended IPO filing lists, nearly all for computing and data centers, naming Broadcom (about $161 billion of equipment leases), Google (about $121 billion) and Amazon ($110 billion) as the largest.
What the S-1/A Says About Chips
Anthropic’s amended S-1, filed on 2 November, uses the phrase “chip design” twice in our search of the text, both in risk factors. One says Anthropic depends “on third-party chip design and services companies for the design, manufacture, and support of certain custom accelerator platforms”.
The other says competitors may be better positioned to vertically integrate “including through proprietary chip design, fabrication” and control of data center infrastructure. The filing adds that Anthropic’s future competitiveness “may depend in part on our ability to successfully develop vertically integrated compute infrastructure on competitive terms and timelines”.
The S-1/A also describes an equipment lease agreed with Broadcom in March 2026 “to procure specialized AI accelerators and associated components”, backed by a note purchase agreement under which Anthropic can issue up to $42.0 billion of convertible notes. We covered the financing in our piece on the $63B SPV for Broadcom chips leased to Anthropic.
The Compute Bill Behind It
The S-1/A says compute and infrastructure expenses “increased sequentially in each of the quarters presented”, driven by the delivery of committed compute under multi-year arrangements. The quarterly table shows $944 million, $1,390 million, $2,047 million and $2,945 million across 2025, then $3,387 million, $5,558 million and $9,863 million in the first three quarters of 2026.
The filing also says the customers paying it $100,000 or more a year included “nine of the 10 largest semiconductor companies” within the Fortune 500 as of 30 September. It does not say what those companies use Claude for.
Connecting the Coverage
Our related pieces on Anthropic’s filing and its chips:
As we read the S-1/A, chips appear as a risk, not a product: the phrase “chip design” occurs twice in our search, once as a dependency on third-party firms and once as an advantage competitors may hold.
As we read the reporting, the time-saving claim is the engineers’ estimate against how they think Nvidia and AMD work, according to The Information’s sources; it is not a benchmark Anthropic has published.
As we read the filing, the stakes are in the cost line: compute and infrastructure expenses rose in every quarter shown, from $944 million to $9,863 million.
Anthropic, amended Form S-1, 2 November 2026
“Our future competitiveness may depend in part on our ability to successfully develop vertically integrated compute infrastructure on competitive terms and timelines”
Three Implications
FOR ANTHROPIC The S-1/A names Broadcom as a partner in a March 2026 lease for specialized AI accelerators, with up to $42.0 billion of convertible notes available to finance it.
FOR NVIDIA AND AMD The Information’s sources compared Claude-assisted design with the engineers’ estimates of workflows at the two companies; neither company is quoted in the free summary.
FOR IPO INVESTORS The Information reports the claim was made in recent investor meetings; the S-1/A excerpts we read do not describe an in-house chip team.
The framework recalls the May 2026 Map of AI, which argued that “the seven-layer AI stack is now governed from the bottom”, naming energy, silicon and compute capacity.
As we read it, a frontier lab using its own model on chip design is a move toward that bottom layer. The framework does not discuss Anthropic’s chip team or Claude’s use in chip design.
What Is Not Established
It is not established how much faster Claude made the team’s work, what the estimate was measured against, which custom chips the team is designing, or whether they have been taped out. The Information’s free summary does not say, and our searches of the amended S-1 for “chip design” and related terms found no description of an in-house chip design team.
We did not find a post on Claude doing chip design on Anthropic’s news, research or engineering pages when we checked their listing pages at 14:16 UTC on 3 November. The 0.3% share is our arithmetic from the reported team size and the filing’s approximate headcount. We did not contact Anthropic.
This Week In The AI Supercycle
Anthropic’s Reality Check, The Capital Squeeze, & The New World Order
This week’s issue: for two years the AI build-out was priced as a growthstory. This week it started being priced as a credit story.
Anthropic has told IPO investors that a 15-person team uses Claude for design work on its custom chips, The Information reports, with engineers estimating time savings against workflows at Nvidia and AMD. The S-1/A frames chips as a dependency and a competitive risk: it relies on third-party chip design firms for custom accelerators, and its compute bill reached $9.86 billion in the third quarter.
95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. This piece is based on The Information’s report of 3 November 2026 (its free summary; the full article is paywalled and we have not read it), a relay of that report on X, and Anthropic’s amended Form S-1 filed with the SEC on 2 November 2026. The 0.3% share is our arithmetic. We did not contact Anthropic.
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.
