Anthropic confirmed on August 5, 2026 that it is building an in-house silicon team to design custom chips for Claude, the first time the AI lab has publicly acknowledged the effort. The confirmation came in a company statement to Business Insider alongside a job listing for a Silicon Engineer that Anthropic posted to its own careers portal.
“We work from the chip level up with our silicon partners, and we are now deepening that investment by building a custom silicon team,” the listing reads. It seeks engineers across front-end design, pre-silicon verification, physical design, design-for-test, analog and mixed-signal, technology and foundry, design infrastructure, and packaging with signal and power integrity, at a salary range of $320,000 to $485,000.
A spokesperson told Business Insider that Anthropic would co-design hardware and models so Claude can run faster and more efficiently “at the scale our customers need,” and said the company has taken and will continue to take a “multi-chip approach” in which hardware from AWS, Google, Nvidia (NVDA ), and AMD remains central to its scaling. That framing matches Anthropic’s own April 6, 2026 compute announcement, where the company said it trains and runs Claude on AWS Trainium, Google TPUs, and NVIDIA GPUs so it “can match workloads to the chips best suited for them.”
A standing ambition, now staffed
The confirmation converts months of signaling into a staffed program. Reuters reported on April 9, 2026 that Anthropic was exploring designing its own chips, though no dedicated team had been committed at that point. By July 2, 2026, TechCrunch reported Anthropic had held talks with Samsung Electronics as a potential manufacturing partner, with the chip’s purpose and specifications not yet decided.
The new listing is the clearest evidence the effort has moved past exploration. It asks for candidates who have “shipped silicon,” can make “consequential calls without a large organization behind them,” and will “support first-silicon bring-up and debug when it arrives.” Those are the responsibilities of a team that expects to tape out a part, not evaluate one.
The economics behind the move are visible in Anthropic’s own disclosures. The company said in April that its run-rate revenue surpassed $30 billion, up from about $9 billion at the end of 2025, and that customers each spending more than $1 million a year had doubled to over 1,000 in under two months. Anthropic’s disclosures tie the revenue surge directly to ‘extraordinary demand from customers worldwide’ that the April compute expansion is meant to serve. Anthropic has been assembling the supply side for some time, including a takeover of SpaceX’s Colossus 1 data center to power Claude earlier this year.
Owning a layer it currently rents
Anthropic’s compute today runs entirely on other companies’ silicon. Its April announcement committed it to multiple gigawatts of next-generation Google TPU capacity coming online starting in 2027, an expansion CFO Krishna Rao called the company’s “most significant compute commitment to date,” and it named Amazon (AMZN ) its primary cloud and training partner through Project Rainier. A custom chip would add a layer Anthropic owns outright alongside that rented capacity.
That is the same logic behind its rivals’ programs. OpenAI unveiled its first custom chip, an inference processor called Jalapeño co-designed with Broadcom (AVGO ), on June 24, 2026. Anthropic’s stated approach is deliberately different in one respect: where OpenAI built a single purpose-built part, Anthropic is explicit that its own silicon will sit inside a multi-vendor stack rather than replace it. The two labs have also explored overlapping supply, with Meta at one point weighing leasing AI compute to Anthropic as demand outpaced available capacity.
What a first Anthropic chip would be optimized for is not yet defined. The listing’s emphasis on machine-learning accelerators, high-bandwidth memory subsystems, on-die interconnect, and hardware-software co-design points at a part built around Claude’s own training and inference patterns. Business Insider notes that designing chips in-house lets AI labs customize their compute to fit their models’ needs and reduce reliance on Nvidia.
Anthropic has not named a manufacturing partner, a process node, or a ship date. The Samsung talks reported in July remain exploratory, and the April TPU expansion means a large share of Anthropic’s near-term capacity is already committed to Google and Broadcom silicon. The custom team is a multi-year bet layered on top of those commitments, and the listing’s search for a small group owning a large surface indicates it is at the start of that build rather than the end.
