Anthropic Considers Developing Custom AI Chips
▼ Summary
– Anthropic is exploring the possibility of designing its own AI chips, though the effort is at an early stage without a committed design or dedicated team.
– The company’s annualized revenue run rate has surpassed $30 billion, creating a scale of compute demand that makes custom silicon worth examining.
– Anthropic recently signed a long-term deal with Google and Broadcom for access to approximately 3.5 gigawatts of TPU-based compute capacity starting in 2027.
– The development cost for an advanced AI chip is estimated at roughly $500 million, a significant but more manageable sum given the company’s rapid revenue growth.
– This exploration mirrors industry moves, as companies like Meta and OpenAI are also working on their own custom AI training chips.
The AI industry’s relentless pursuit of computational power is pushing leading firms to consider a fundamental shift in strategy. Anthropic, the company behind the Claude AI models, is now exploring the development of its own custom artificial intelligence chips. According to a recent report, this internal effort remains in a preliminary phase, with no final decision made on whether to proceed with a proprietary design or continue relying on external suppliers.
This strategic consideration follows a period of explosive growth for the company. Earlier this week, Anthropic disclosed that its annualized revenue run rate has surged past $30 billion, a dramatic increase from roughly $9 billion just four months prior at the end of 2025. This rapid scaling has created a massive demand for computing resources, making the long-term investment in custom silicon a more compelling economic proposition. Currently, Anthropic operates its models across a diversified portfolio of hardware, including Google’s Tensor Processing Units, Amazon’s custom chips, and Nvidia GPUs, selecting the optimal processor for each specific workload.
The timing of this exploration is notable, coming just days after the company solidified a major partnership. Anthropic signed a long-term agreement with Google and Broadcom securing access to approximately 3.5 gigawatts of TPU-based compute capacity starting in 2027. This represents a significant expansion, roughly triple the one gigawatt of power the company was consuming earlier in 2026. A regulatory filing from Broadcom included an unusual condition, noting that this expanded deployment is contingent on Anthropic’s continued commercial success. This deal builds upon a previous $50 billion commitment from Anthropic to invest in U. S. computing infrastructure made in November 2025.
Broadcom’s role is particularly significant, as the firm is already a chip design partner for OpenAI and has other undisclosed clients in the AI space, positioning it centrally in the emerging market for specialized AI processors. The potential move by Anthropic reflects a broader industry trend toward vertical integration in hardware. Competitors like Meta have been developing their own AI training chips, and OpenAI is also known to be working on custom silicon projects.
The barrier to entry, however, is substantial. Industry sources estimate the development cost for an advanced AI chip at around $500 million, a sum that covers specialized engineering talent and manufacturing validation. While this figure is significant for a company that is not yet profitable, it appears more manageable when weighed against a revenue base that has more than tripled in a matter of months. For now, the company maintains its options, weighing the immense upfront investment against the potential for greater efficiency and control over its core technological infrastructure.
(Source: The Next Web)
