Anthropic is considering the development of its own AI chips.

      Anthropic is in the early stages of exploring the design of its own AI chips, but it may opt to continue purchasing chips instead of creating its own. This exploration takes place shortly after the company entered a long-term agreement with Google and Broadcom for 3.5 gigawatts of TPU computing power starting in 2027. A representative from the company chose not to comment on the matter.

      According to a report from Reuters on Thursday, Anthropic is looking into the possibility of developing its own AI chips, citing three sources with knowledge of the situation. However, as of now, there is no commitment to a specific design or a dedicated team for this initiative. The company still might prefer to buy chips from external sources rather than manufacture its own. A spokesperson from the San Francisco-based firm did not comment on the report.

      This exploration comes as Anthropic has seen a significant increase in its revenue. Earlier this week, the company revealed that its annualized revenue run rate has exceeded $30 billion, up from about $9 billion at the end of 2025. This growth has led to a scale of computing demand that makes the prospect of custom silicon more economically viable. Currently, Anthropic operates Claude using a combination of chips, including tensor processing units developed by Google in collaboration with Broadcom, as well as Amazon's custom chips and Nvidia hardware. The company adjusts workloads to the chips that are most appropriate for them.

      Just days prior to the Reuters report, Anthropic finalized a long-term agreement with Google and Broadcom for approximately 3.5 gigawatts of TPU-based computing capacity from 2027, which is about three times the one gigawatt it was using earlier in 2026, as noted in Broadcom’s SEC filing. This filing mentioned that the increased deployment is dependent on Anthropic's ongoing commercial success, which is an unusual caveat for a regulatory document. The deal builds upon Anthropic's commitment made in November 2025 to invest $50 billion in computing infrastructure in the U.S.

      Broadcom is already a chip design partner for OpenAI and has another undisclosed XPU client, positioning itself prominently in the emerging market for custom AI silicon as an alternative to Nvidia's general-purpose GPUs.

      Anthropic's potential development of proprietary silicon reflects similar initiatives across the industry. Meta has been creating its own AI training chips, while OpenAI is also working on custom silicon. Industry sources mentioned by Reuters estimate the development cost of an advanced AI chip to be around $500 million, accounting for the necessity of hiring specialized engineers and validating the manufacturing processes. While this amount is substantial for a company that remains unprofitable, it becomes more feasible in light of a revenue run rate that has more than tripled in just four months.

      Published April 10, 2026 - 7:50 am UTC