Anthropic is assembling a team to design its own custom chips for AI, according to a Business Insider report, as it seeks to co-design hardware and models so its Claude systems run faster and more efficiently. The move reflects surging demand for Claude and intensifying competition among AI firms for scarce infrastructure, suggesting that relying solely on external hardware partners is no longer sufficient to meet Anthropic's scaling needs.

Anthropic currently sources computing hardware through deals with AWS, Google, Nvidia and AMD, and last month was reported to be scouting Samsung as a potential manufacturing partner for custom chips. It joins other major AI players moving in-house on silicon: OpenAI unveiled its Broadcom-built Jalapeño inference chip in June, Google DeepMind has long used Alphabet's TPUs, and Meta has been developing its own MTIA accelerators. A job listing shows Anthropic is now recruiting engineers for a "custom silicon team"; the company did not respond to a request for comment.