Anthropic Explores Building Its Own AI Chips as Revenue Tops $30 Billion

Anthropic is exploring the possibility of designing its own AI chips, according to a Reuters exclusive published today, marking the latest move by a frontier AI lab to secure greater control over its computing infrastructure. The plans are still in early stages — no dedicated chip team has been assembled and no specific design has been committed to — but the initiative signals a strategic shift for a company that has relied heavily on Google TPUs and Amazon's Trainium processors.

Why Custom Chips Now?

The timing aligns with Anthropic's extraordinary growth trajectory. The company's annualized revenue run rate has surpassed $30 billion in 2026, tripling from roughly $9 billion at the end of 2025. With that scale of demand for inference and training compute, even small efficiency gains from purpose-built silicon could translate into billions of dollars in savings.

Developing advanced AI chips is not cheap — industry estimates put the cost at roughly $500 million for a competitive custom chip program. But for a company generating revenue at Anthropic's pace, the economics increasingly favor in-house development over perpetual reliance on external suppliers.

A Growing Trend Among AI Labs

Anthropic would be joining a growing cohort of AI companies pursuing custom silicon. Meta has been designing its own training and inference chips for several years, while OpenAI has explored custom chip development as part of its broader infrastructure strategy. Google, of course, has long operated its own TPU program, which Anthropic has been one of the largest external consumers of.

The move also comes days after Anthropic signed a multi-year GPU cloud deal with CoreWeave and its landmark 3.5-gigawatt infrastructure agreement with Google and Broadcom. Together, these deals give Anthropic access to multiple compute architectures — Nvidia GPUs via CoreWeave, Google TPUs, Amazon Trainium chips, and potentially its own custom silicon in the future.

Diversification as Strategy

The chip exploration reflects a broader diversification strategy at Anthropic. Rather than betting on a single compute provider, the company is building optionality across the entire hardware stack. This approach reduces supply chain risk at a time when demand for AI compute far outstrips available capacity.

An Anthropic spokesperson declined to comment on the Reuters report. The company has previously emphasized the importance of compute access to its mission of building safe, beneficial AI systems.

What It Means for the Industry

If Anthropic proceeds with a custom chip program, it would further blur the line between AI software companies and hardware infrastructure providers. The trend suggests that at the frontier of AI development, the traditional separation between model developers and chip designers is breaking down — driven by the enormous scale of compute required for next-generation AI systems and the competitive advantage that custom silicon can provide.