Microsoft’s Maia AI chips could soon power Anthropic’s Claude models

Microsoft is reportedly in discussions with AI start-up Anthropic over a possible agreement to use Microsoft’s custom-built artificial intelligence chips, a move that could strengthen the software company’s position in the increasingly competitive AI infrastructure market.

According to a CNBC report, the talks involve Microsoft’s Maia AI processors, which were introduced as part of the company’s push to reduce dependence on third-party chip suppliers such as Nvidia. While Microsoft unveiled the second-generation Maia 200 chips earlier this year, the processors have not yet been widely offered through the company’s Azure cloud platform.

The potential partnership would represent a notable step forward for Microsoft, which has trailed rivals Amazon and Google in securing customers for specialised AI hardware.

Anthropic has not yet finalised an agreement with Microsoft, and negotiations remain ongoing. Neither company has publicly confirmed the discussions, reports CNBC.

The talks come at a time when Anthropic is rapidly expanding and facing mounting pressure to secure enough computing resources to support its growing AI services. The company’s Claude chatbot and its Claude Code programming assistant have gained significant traction this year, driving up demand for advanced processing infrastructure.

More from Tech

Earlier this month, Anthropic chief executive and co-founder Dario Amodei acknowledged the company had encountered challenges related to computing capacity as user demand accelerated.

Anthropic has traditionally relied heavily on Nvidia’s graphics processing units to train and operate its generative AI models. However, the company has increasingly diversified its hardware partnerships in an effort to secure long-term access to AI computing power.

In April, Anthropic announced a decade-long arrangement with Amazon Web Services to use the cloud provider’s custom Trainium chips in a deal reportedly valued at more than $100 billion. Last year, the company also confirmed plans to use Google’s tensor processing units for some of its AI workloads.

The growing competition for AI infrastructure has intensified as technology firms race to develop alternatives to Nvidia’s dominant chips, which remain in extremely high demand across the industry.

Microsoft has been aggressively investing in AI infrastructure and strategic partnerships. In November, the company revealed plans to invest $5 billion in Anthropic, while the AI firm committed to spending around $30 billion on Microsoft’s Azure cloud services over time.

Meanwhile, Anthropic’s appetite for computing power continues to expand. This week, SpaceX disclosed that Anthropic had agreed to pay $1.25 billion per month through 2029 for additional computing capacity.

Microsoft chief executive Satya Nadella said during the company’s April earnings call that the Maia 200 chips deliver more than 30 per cent better “tokens per dollar” performance compared with other processors currently used in Microsoft’s infrastructure. He added that the chips are already operating in Microsoft data centres in Arizona and Iowa.