Anthropic confirmed on Aug. 5, 2026, that it is forming an in-house custom-silicon team to design chips for Claude training and inference, targeting lower operating costs and response latency. The company reportedly expects to spend about $19 billion on compute this year and has hired former OpenAI engineer Clive Chan, who joined in early June, to help lead the effort. Relevant engineering roles offer salaries of $320,000 to $485,000. Anthropic has held exploratory talks with Samsung Electronics since July about potential chip manufacturing, while continuing its multi-chip strategy with Nvidia, AMD, AWS Trainium and Google TPUs. Google is expanding next-generation TPU capacity by about 3.5 gigawatts from 2027, with Broadcom facilitating the buildout, and Anthropic is expected to use that capacity before its own silicon matures. The project follows similar efforts by OpenAI, Meta, Google, Amazon and Microsoft and could eventually reduce Anthropic's reliance on cloud and chip suppliers. Deployment has previously been discussed as taking 18 months to three years, implying a potential market launch around 2028, although advanced packaging, high-bandwidth memory and foundry capacity remain risks. Anthropic's rapid growth—including roughly $11.6 billion in second-quarter 2026 revenue, an annualized run rate above $65 billion by the end of July and an estimated valuation of $965 billion—has increased the strategic importance of controlling compute costs.