Anthropic confirms in-house AI chip team
Anthropic publicly confirmed it is building custom chips for Claude, hiring senior chip engineers at salaries up to $485,000 as part of a multi-chip strategy.
Anthropic confirmed on August 5 that it is building an in-house team to design custom silicon for its Claude models, ending months of reporting about talks with outside chipmakers by putting a public job listing behind the plan. The postings seek senior engineers who have personally shipped finished semiconductor designs, offering salaries up to $485,000, and ask candidates to be comfortable making high-stakes calls without a large supporting organization yet in place.
Anthropic frames this as a “multi-chip” strategy that adds to, rather than replaces, its existing use of Nvidia, AMD, Google, and AWS hardware. The company has not named a manufacturing partner, given a timeline, or said whether it will fabricate chips itself; Samsung has been floated as a possible manufacturing partner in earlier reporting about custom-chip talks, but nothing has been confirmed since. The stated goal is software-hardware co-design: shaping chips and models together to cut inference cost and latency at scale, the same argument OpenAI made when it revealed its Broadcom-built Jalapeño chip and that Google and Amazon have made for years with TPUs and Trainium.
What it means for you
Nothing changes for API users today. What this confirms is that every major model lab now considers chip supply too important to leave entirely to Nvidia, and Anthropic just moved from “reportedly talking to a manufacturer” to “publicly building a design team,” which is a multi-year commitment, not a pilot. If Claude pricing or capacity gets materially more stable relative to competitors over the next two to three years, in-house silicon co-designed with the models is one of the reasons why.