Anthropic Exploring Custom AI Chips to Fuel Claude’s Growth

SAN FRANCISCO – Anthropic, the AI research lab behind the popular Claude chatbot, is reportedly exploring the development of its own custom AI hardware. According to sources familiar with the matter, the startup is in the early stages of weighing whether to design in-house chips to mitigate the ongoing global shortage of high-end semiconductors.

While the plans remain preliminary—with no dedicated team yet assigned—the move signals a potential shift in how the company manages its massive scaling requirements.

Scaling Amidst a Chip Crunch

The internal discussions come at a time of explosive growth for the San Francisco-based firm. Earlier this week, Anthropic revealed that its run-rate revenue has skyrocketed to over $30 billion in 2026, a massive jump from the $9 billion reported at the end of 2025.

Currently, Anthropic relies heavily on infrastructure from its major backers, utilizing:

• Google’s Tensor Processing Units (TPUs)

• Amazon’s custom AI chips

• Broadcom-assisted designs

A recent long-term agreement with Google and Broadcom underscores Anthropic's commitment to U.S. computing infrastructure, involving a projected $50 billion investment. However, designing bespoke silicon could allow the company to optimize hardware specifically for the architecture of its Claude models, potentially increasing efficiency and lowering long-term operational costs.

Joining the Silicon Arms Race

If Anthropic proceeds, it will join an elite group of "hyperscalers" and AI rivals—including OpenAI, Meta, and Google—who are all racing to reduce their dependency on third-party chip manufacturers like Nvidia.

The financial stakes of such a venture are high. Industry experts estimate that bringing a single advanced AI chip to market can cost upwards of $500 million, factoring in specialized engineering talent and complex manufacturing processes.

"The plans are in early stages and the company may still decide to only buy AI chips and not design any," noted one source briefed on the discussions.

As of now, Anthropic has declined to comment on the report. Whether the company builds its own "brain" or continues to lease power from tech giants, the move highlights the intensifying battle to secure the raw computational power necessary to lead the next generation of artificial intelligence.