Anthropic has confirmed plans to build an internal "custom silicon team" to design the chips required to run its AI models. The confirmation followed the discovery of job listings for senior semiconductor engineers and technical program managers on the company’s career board.
A Multi-Chip Strategy
A spokesperson for Anthropic clarified that the company will maintain a "multi-chip approach," intending to use hardware from external providers alongside its in-house designs as it scales. This move validates earlier reports suggesting Anthropic was exploring a manufacturing partnership with Samsung.
Escaping Nvidia’s Gravity
Anthropic follows a growing industry trend of vertical integration. Competitor OpenAI recently unveiled its "Jalapeño" chip, developed with Broadcom, while tech giants like Google and Meta have already deployed proprietary silicon. Mistral is also reportedly considering a similar move. The shift is driven by several factors:
- Reducing strategic vulnerability caused by heavy reliance on Nvidia hardware.
- Addressing a competitive environment where compute demand outstrips supply.
- Optimizing performance by co-designing hardware and models side-by-side.
The company believes this integration could provide a competitive edge as the market explores smaller or open-weight models. However, since the team is still being recruited, the actual deployment of Anthropic’s custom silicon remains a long-term goal.
