Anthropic Explores Building Its Own AI Chips as Revenue Surges
April 10, 2026 – 7:50 am
Anthropic is considering designing its own AI chips, as reported by Reuters on Thursday, based on sources familiar with the matter. Although in an early stage, the company hasn’t committed to a specific design and has yet to form a dedicated team for this project. They may still opt to continue purchasing chips from third parties instead.
A spokesperson for Anthropic refused to comment on the news. This exploration comes amid rapid revenue growth; Anthropic’s annualized run rate exceeded $30 billion, up from approximately $9 billion at the end of 2025.
The company currently operates Claude on a variety of chips, including:
- Tensor processing units (TPUs) designed by Alphabet’s Google in partnership with Broadcom
- Amazon’s custom chips
- Nvidia hardware
Anthropic matches workloads to the most suitable chip for each task.
Just days prior to the Reuters report, Anthropic secured a long-term deal with Google and Broadcom granting them access to approximately 3.5 gigawatts of TPU compute capacity starting in 2027—roughly three times their previous consumption in 2026, as revealed in Broadcom’s SEC filing.
This expanded deployment depends on Anthropic’s continued success, an unusual condition for a regulatory document. This deal builds upon their November 2025 commitment to invest $50 billion in US computing infrastructure.
Broadcom, already a chip design partner for OpenAI and another undisclosed XPU customer, is at the forefront of the emerging custom AI silicon market.
Other industry players are also making moves in this direction; Meta has been developing its own AI training chips, while OpenAI has been working on custom silicon as well. Industry sources tell Reuters that developing an advanced AI chip costs around $500 million—a significant investment for a company currently unprofitable but manageable against a rapidly growing revenue base.