Anthropic has confirmed it is building an in-house silicon team to co-design custom AI chips alongside its Claude models, joining OpenAI and Google in the race to control the full stack of AI hardware. The company disclosed the plan on August 5, 2026, and is actively hiring semiconductor engineers with salaries reaching up to $485,000.
A co-design bet aimed at Claude economics
Anthropic's pitch to the hardware community is unusually explicit: build engineers who can work fluidly between chip architecture and model architecture, so that Claude's next generations can be tuned to silicon that is designed for them. Executives have privately argued the approach could roughly halve inference cost per token at production scale — a huge lever given the company's revenue and compute demand curves.
Clive Chan anchors the new team
The nascent silicon effort is anchored by Clive Chan, who joined Anthropic in early June 2026 from Tesla's Dojo program, where he was one of the earliest hardware hires. Chan previously spent time as the second hardware hire on OpenAI's dedicated chip team. Anthropic is now seeking additional physical-design, verification and infrastructure engineers.
Compute demand exploding
Anthropic's run-rate revenue has surged from roughly $9 billion at the end of 2025 to more than $30 billion in mid-2026, and the company recently expanded its multi-gigawatt Google Cloud and Broadcom partnership to lock in additional TPU capacity. Custom silicon is a natural next step — and one that echoes the strategic logic behind Nvidia's $500 billion compute financing platforms.
What it does — and doesn't — mean
Anthropic is explicit that any Anthropic-designed accelerator is at least several years from production; the team's current phase is closer to research and architecture than tape-out. The move should be read alongside River AI's $1.1B stealth raise and other billion-dollar AI stack bets as evidence that the frontier is now competing on hardware, model and product architecture simultaneously.
Reporting based on coverage from TechCrunch, Reuters and Forbes.