Anthropic is assembling an in-house team to develop custom AI chips for Claude, according to Forbes, which reported that the move puts the company alongside OpenAI, Meta and Google in the race to build proprietary AI silicon.
The Ukrainian outlet Mezha reported the same development, saying Claude is set to get its own AI chips and that Anthropic is already putting together a team of developers to make it happen. A report from indiagazette.com framed the effort as a response to hardware shortages, noting Anthropic is building a custom AI chip team amid constrained supply.
Custom silicon is the industry's current answer to a hard problem: the specialized processors that train and run large AI models are scarce, expensive, and largely sourced from a small number of suppliers. Designing your own chip is slow and capital-intensive, but it can reduce dependence on outside vendors and let a company shape the hardware around the specific demands of its own models.
Until now, Anthropic had been the notable holdout among the largest AI labs on this front. Google has long built its own AI accelerators, and both OpenAI and Meta have pursued custom designs. Anthropic staffing up its own chip team means every major frontier lab is now trying to control some part of the physical layer beneath its software.
The reports describe a team being formed, not a finished product, and none of the sources give a timeline, a manufacturing partner, or technical specifications for what Anthropic might build.
Why it matters: if the companies building the most capable AI models also start building the chips those models run on, the economics and the bottlenecks of the entire AI industry shift — affecting who can afford to compete and how quickly AI capabilities reach everyone else.