Anthropic in talks to rent Microsoft AI chips amid rising demand for compute

Anthropic is in early-stage discussions to rent Microsoft Azure servers equipped with the tech giant’s custom Maia 200 AI chips, a deal that would mark a meaningful win for Microsoft’s fledgling chip program and give Anthropic fresh compute muscle for its Claude models.

The talks, first reported by The Information, have not yet produced a formal agreement.

What’s on the table

Microsoft launched the Maia 200 chips in January 2026, touting over 30% improvement in tokens processed per dollar compared to previous models. The chips are currently operational in Microsoft’s data centers in Arizona and Iowa. If a deal materializes, Anthropic would be accessing that capacity through Azure’s cloud infrastructure rather than purchasing hardware outright.

Anthropic already committed a minimum of $30 billion to Azure compute back in November 2025. That commitment leaned heavily on NVIDIA technologies at the time. Layering in Microsoft’s own custom silicon would represent a meaningful evolution of the relationship.

For Microsoft, the Maia chip program hasn’t seen the kind of external adoption that would validate the investment. Landing Anthropic as a customer for custom silicon would be a significant credibility boost.

Why Anthropic needs more chips

Anthropic CEO Dario Amodei flagged earlier this month that the company faces significant compute challenges.

Anthropic’s answer has been diversification. Beyond the Microsoft relationship, Anthropic has a decade-long engagement with Amazon involving Trainium chips and has previously accessed Google Cloud’s TPUs.

The bigger picture for AI hardware

Amazon has its Trainium and Inferentia chips. Google has TPUs that have been in production for years. Microsoft’s Maia line is the relative newcomer, and it needs reference customers to prove the chips work at scale for external workloads.

The inference market, where trained models actually serve users, is increasingly where custom chips can compete on cost efficiency. That’s exactly the use case Anthropic would be exploring with the Maia 200.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.