What happened
Anthropic is discussing custom AI chip manufacturing with Samsung Electronics, though the project remains in early stages without a finalised design. Anthropic hired Clive Chan, who previously developed OpenAI's custom chip program, signalling a move from exploration to active development. This follows OpenAI's recent unveiling of its Broadcom-built "Intelligence Processor" custom inference chip. Anthropic's annualized revenue run rate surpassed $30 billion this year, more than tripling from roughly $9 billion at the end of 2025, and it secured a long-term deal with Google and Broadcom for roughly 3.5 gigawatts of TPU compute starting 2027.
Why it matters
Control over hardware supply chains and compute costs is shifting for AI developers. This move by Anthropic, following OpenAI's custom chip launch, indicates a broader industry drive to reduce reliance on single-vendor hardware, particularly Nvidia. For platform engineers and procurement teams, this signals a future where access to frontier models increasingly depends on diversified silicon strategies, potentially locking in long-term compute costs and performance. The shift prioritises hardware control, impacting model deployment flexibility and operational expenditure.
