Anthropic is in early talks to rent servers powered by Microsoft’s Maia AI chips, The Information reported Thursday, citing two people familiar with the discussions. The talks would give the Claude maker another option for running inference workloads as demand for its AI services rises. The discussions may still end without a deal, but they would put Microsoft’s custom silicon inside a compute strategy already split across Amazon, Google and Nvidia.

Key Takeaways

AI-generated summary, reviewed by an editor. More on our AI guidelines.

Talks are early

Microsoft introduced Maia 200 on Jan. 26 as an inference accelerator for Azure. A possible rental would sit beside Anthropic’s November Azure deal with Microsoft and Nvidia. Microsoft’s post put Maia 200 on Taiwan Semiconductor Manufacturing Co.’s 3-nanometer process and described it as a chip for serving already trained models, not for the full training runs covered by Anthropic’s other supply agreements.

The reports describe the talks as a possible Maia-powered server rental for Claude workloads; they do not indicate any change to Anthropic’s existing training arrangements. Reuters and Bloomberg reported that the talks are early; Reuters said Microsoft and Anthropic did not immediately respond to requests for comment.

Azure already has Claude

Anthropic’s Microsoft relationship already includes the November partnership with Microsoft and Nvidia. Under that agreement, Anthropic committed to purchase $30 billion of Azure compute capacity and to contract up to 1 gigawatt of additional compute capacity, initially using Nvidia Grace Blackwell and Vera Rubin systems.

Microsoft and Nvidia also committed to invest up to $15 billion in Anthropic. The November launch gave Microsoft Foundry customers access to Claude Sonnet 4.5, Claude Opus 4.1 and Claude Haiku 4.5; Microsoft has since added newer Claude releases, including Opus 4.6.

Anthropic spreads supply

Anthropic has separate supply arrangements outside Azure. The company announced an Amazon agreement on April 20 that commits more than $100 billion to Amazon Web Services (AWS) technologies over 10 years and reserves up to 5 gigawatts using Trainium and Graviton chips.

Anthropic said the Amazon deal includes more than 1 million Trainium2 chips already in use and nearly 1 gigawatt of Trainium2 and Trainium3 deployments by the end of 2026. On April 6, it also announced a Google and Broadcom agreement for multiple gigawatts of tensor processing unit (TPU) supply expected to come online starting in 2027, adding to an earlier Google cloud pact.

Microsoft still has to prove scale

Microsoft says Maia 200 is deployed in US Central and will serve GPT-5.2 models, Microsoft Foundry, Microsoft 365 Copilot and the Microsoft Superintelligence team. Bloomberg noted that Microsoft began custom-chip work later than Google and Amazon, and that its chips are viewed as less mature and available in smaller quantities than rival cloud accelerators.

Microsoft says Maia 200 is already deployed in the US Central Azure region near Des Moines, Iowa, with US West 3 near Phoenix next and other regions to follow.

If completed, a Maia rental agreement would give Microsoft a named frontier-model customer for its own silicon. Until either company confirms the discussions, the confirmed company record consists of Microsoft’s Maia launch post and Anthropic’s existing Azure, AWS and Google/Broadcom supply agreements. No Azure document yet names Maia in a Claude workload.

Meta Turns to Amazon Graviton Chips as AI Agents Shift Compute Math
Meta has signed a multiyear deal to run AI workloads on Amazon's Graviton processors, giving AWS one of its largest outside validations for homegrown server chips. The deployment begins with tens of m
The Implicator
Tesla and SpaceX Pitch $25B Terafab Chip Project Starting in Austin, Offer No Timeline
Elon Musk announced Saturday that Tesla and SpaceX will pursue a $20 to $25 billion chip fabrication project beginning with an advanced technology facility in Austin, Texas, Reuters reported. The proj
The Implicator
Nvidia Didn't Just Launch Chips at GTC. It Launched a Lock-In Machine.
Monday at the SAP Center in San Jose, Jensen Huang held up a chip. Rotated it under the stage lights, slow, deliberate, the way he always does. A jeweler showing off a diamond. Thirty thousand people
The Implicator

Frequently Asked Questions

What are Anthropic and Microsoft discussing?

Anthropic is reportedly in early talks to rent servers powered by Microsoft Maia AI chips for Claude workloads. Reuters and Bloomberg said the talks may not lead to an agreement.

What is Maia 200?

Maia 200 is Microsoft’s Azure inference accelerator, introduced Jan. 26. Microsoft says it uses TSMC’s 3nm process, HBM3e memory and on-chip SRAM for model-serving workloads.

Does this change Anthropic’s AWS or Google deals?

The reports do not indicate a change. Anthropic still has a $100 billion-plus AWS commitment and a Google-Broadcom TPU supply agreement expected to start in 2027.

How is Microsoft already connected to Anthropic?

Anthropic’s November deal included $30 billion of Azure compute capacity, up to 1 gigawatt of additional Nvidia-based infrastructure and broader Claude access in Microsoft Foundry.

Why would a Maia deal matter for Microsoft?

A completed rental would give Microsoft a named frontier-model customer for Maia. Microsoft still has to show its custom silicon can scale beyond its own workloads.

AI-generated summary, reviewed by an editor. More on our AI guidelines.

AI News

San Francisco

Editor-in-Chief and founder of Implicator.ai. Former ARD correspondent and senior broadcast journalist with 10+ years covering tech. Writes daily briefings on policy and market developments. Based in San Francisco. E-mail: [email protected]