Anthropic in talks with Samsung to develop a custom AI inference chip

Anthropic, the developer of the Claude language model, is in early-stage negotiations with Samsung Electronics to design and manufacture its own AI inference chip. The talks focus on using Samsung’s advanced 2‑nanometer foundry process and cutting‑edge packaging technology to create a silicon solution tailored to Anthropic’s models, aiming to reduce reliance on Nvidia GPUs and lower operating costs.

The move follows OpenAI’s recent launch of the “Jalapeño” chip built with Broadcom and reflects a broader industry trend of AI firms seeking bespoke hardware. Anthropic has hired Clive Chan, a former member of OpenAI’s custom‑chip team, to lead the effort and counted Samsung, SK hynix and Micron as strategic infrastructure partners after they invested in Anthropic’s Series H funding round.

Anthropic stresses it will continue using existing providers—Nvidia GPUs, Google TPUs and Amazon’s Trainium—while the Samsung partnership could give the Korean chipmaker a high‑profile AI customer and help narrow the gap with market leader TSMC. News of the potential deal lifted Samsung’s shares about 8 % on the Seoul exchange.