Anthropic to develop its own AI chips, forming a custom silicon team

Anthropic, the company behind the Claude AI models, is now planning to develop its own AI chips to reduce dependencies and increase control over costs. According to TechCrunch, the company is seeking hardware and software experts for its new 'Custom Silicon Team', with candidates expected to bring chip design expertise and contribute to developing new AI models. Anthropic aims to make its AI technology faster and more efficient. In July 2026, reports had surfaced that Anthropic was in negotiations with Samsung, which would manufacture a chip designed by the AI firm. Meanwhile, OpenAI uses Broadcom for the production of its Jalapeño chip, designed for inference with lower latency and energy consumption. Anthropic is also collaborating with Broadcom and Google on special Tensor Processing Units, and Google agreed in April to invest up to $40 billion in Anthropic. The company has contracts with Amazon (AWS) and Alphabet (Google Cloud Platform) for AI infrastructure, and with Nvidia and AMD for hardware. According to Reuters, Anthropic sees the custom chip team as a complement to its existing infrastructure, maintaining a multi-chip strategy. The timeline for the first Anthropic chip remains unknown, but industry experts cited by Reuters estimate that designing a modern AI chip alone could cost around $500 million.