Anthropic joins AI giants building in-house chips as demand outpaces supply

Anthropic just confirmed today that it has begun assembling its own dedicated team to design custom chips for its AI, Claude.

Until now, the company relied entirely on chips made by other tech giants, but skyrocketing demand means they need their own hardware to keep things running fast and smooth.

A job listing turns months of rumors into a staffed program

The news came via an official statement to Business Insider, where they also revealed a new opening for a lead silicon engineer. According to Business Insider, an Anthropic spokesperson said the company would co-design its hardware and its models so that Claude runs “faster and more efficiently,” in the spokesperson’s words, “at the scale our customers need.”

Interestingly, the job description is quite strict. Apparently, Anthropic is looking for engineers across front-end design, pre-silicon verification, physical design, analog and mixed-signal work, and packaging, who can work within a salary band of $320,000 to $485,000. 

According to the listing, candidates must show “direct personal contribution” to finished, shipped semiconductor designs, and be ready to make “consequential calls without a large organization behind them.”

The listing also asks the new hire to “support first-silicon bring-up and debug” once a part arrives, the kind of language that describes a team expecting to tape out a chip rather than evaluate someone else’s. 

While outlets like Reuters and The Information previously hinted at talks with manufacturing partners like Samsung, Anthropic hasn’t named a factory or a release date yet.

Owning the layer it currently rents

Anthropic also made it clear that they weren’t ditching their current partners. The spokesperson stressed the tech giant’s plan to keep using chips from AWS, Google, Nvidia, and AMD alongside their new hardware to keep their supply chain flexible.

However, the economics behind the decision lie in Anthropic’s own numbers. According to reports, the company said in April that its run-rate revenue had passed $30 billion, up from roughly $9 billion at the close of 2025, and that customers spending more than $1 million a year had doubled to over 1,000. 

All of that compute currently runs on other companies’ silicon: AWS Trainium, Google TPUs, and Nvidia GPUs, with multiple gigawatts of next-generation Google TPU capacity committed from 2027. A chip Anthropic owns would give it one layer of that stack it no longer has to rent.

OpenAI and Meta got there first

Anthropic is not the first frontier lab to make this bet. Its closest rival, OpenAI, unveiled a purpose-built inference chip called Jalapeño, developed with Broadcom, in June. OpenAI claims the part can cut costs by up to 50%.

Other competitors are moving even faster. Meta is gearing up to start making its own chips soon, and even smaller startups are looking into the idea. The goal for everyone is the same: stop relying so heavily on Nvidia, whose chips currently dominate the market, and build hardware tailored specifically to their own AI.

For Anthropic, this also helps with their recruitment. While there are concerns that some workers care more about big tech paychecks than the company’s core mission, Anthropic, now offering the chance to build brand-new hardware from the ground up, gives top-tier engineers a brand-new reason to sign on.

If you're reading this, you’re already ahead. Stay there with our newsletter.