AI lab Anthropic is said to be in talks to use Microsoft's custom AI chips.

As reported by The Information and citing executives involved in the discussion, the possibility of using Microsoft's custom chips comes as the AI lab continues to increase its Azure cloud usage.

Microsoft is behind the Maia AI chips, and earlier this year, the company announced that the latest generation - the Maia 200 - is now running in its data centers.

The Maia 200 is built using TSMC's 3nm technology and delivers around 10 petaflops of FP4 and five petaflops of FP8 compute performance within a 750W SoC (System-on-Chip) TDP (Thermal Design Power) envelope, the chip has a “redesigned memory system” comprising 216GB HBM3e at seven Tbps and 272MB of on-chip SRAM, alongside data movement engines.

The cloud giant claims that it is more performant than the latest generation of AI chips from its hyperscaler competitors, Amazon's Trainium and Google's TPUs.

Microsoft and Anthropic have been working together since November 2025, when Anthropic agreed to purchase $30 billion of Microsoft Azure compute capacity, while Microsoft said it would invest $5bn in the AI lab.

Anthropic has been rapidly expanding its compute capacity, and has long been open to adopting a variety of hardware to train its models.

In recent months, the company has signed on to lease the xAI data center from SpaceX, which comes with a monthly bill of $1.25bn, signed a $1.8bn contract with Akamai, and has reportedly been in discussions to purchase inference chips from a UK startup - Fractile.

On the hyperscaler side, Anthropic has long used Google Cloud for some of its model work, and in October 2025, the company announced a 1GW+ cloud deal with up to 1 million Google TPUs.

The generative AI company separately announced a $50bn data center investment at Fluidstack facilities – although they are the same sites Google operated out of, so it may be the same deal.

Amazon is also an Anthropic investor, pumping $8bn into the company. In October, AWS launched the Project Rainier cluster featuring its own Trainium2 chips, for use by the AI firm. It plans to spend as much as $11bn on the site at full buildout.