Anthropic mulls custom AI chips
Reuters reports Anthropic is exploring the possibility of building its own AI chips to reduce reliance on scarce GPUs, a step the company is considering as enterprise revenue scales. Coverage frames the move as a response to chronic GPU shortages and as a signal that vendors are weighing vertical integration to secure capacity. (cnbc.com, thenextweb.com)
Anthropic is exploring whether to design its own artificial intelligence chips instead of relying only on outside suppliers, according to Reuters. (cnbc.com) Reuters reported on April 10 that the talks are still early and Anthropic could decide not to build chips at all. The report said the company is weighing the move as shortages of artificial intelligence processors continue to constrain model training and deployment. (cnbc.com) These chips are the specialized processors that run the heavy math behind systems like Claude. Today Anthropic says it trains and serves Claude across Amazon Web Services Trainium, Google tensor processing units, and Nvidia graphics processing units. (anthropic.com, aboutamazon.com) Anthropic has spent the past year locking in outside capacity instead of owning the silicon itself. On April 6, the company said it signed a new agreement with Google and Broadcom for multiple gigawatts of next-generation tensor processing unit capacity starting in 2027. (anthropic.com) Amazon is also deeply tied to Anthropic’s infrastructure. In November 2024, Anthropic said Amazon would invest another $4 billion, bringing Amazon’s total investment to $8 billion, and named Amazon Web Services its primary cloud and training partner. (anthropic.com, aboutamazon.com) The chip question is surfacing as Anthropic’s business expands fast. Anthropic said in March 2025 that it raised $3.5 billion at a $61.5 billion post-money valuation, and recent coverage said Claude’s annualized revenue run rate has climbed past $30 billion. (anthropic.com, thenextweb.com) Building a custom chip would push Anthropic further into vertical integration, where a company controls more of the stack from model software to computing hardware. That approach can lower long-term costs and secure supply, but it also demands years of design work, manufacturing partners, and large capital commitments. (cnbc.com, anthropic.com) Anthropic has not publicly announced a chip program of its own, and its current posture remains multi-supplier rather than single-stack. For now, the company is still buying time and capacity from Amazon, Google, Broadcom, and Nvidia while it decides whether to go one layer deeper. (cnbc.com, anthropic.com, aboutamazon.com)
