Microsoft in talks with Anthropic

Microsoft and Anthropic are in talks about giving the startup access to Microsoft’s Maia artificial-intelligence chips, according to CNBC and Bloomberg, adding a new hardware layer to a relationship the companies expanded in November 2025. Anthropic and Microsoft have not announced a chip deal, and CNBC reported on May 21 that no agreement had been signed. The talks come as Anthropic adds cloud and chip suppliers to support rising demand for Claude, its AI models and coding tools. They also arrive as investors track how a small group of private AI companies are drawing unusually large sums of capital and infrastructure commitments. ### Which Microsoft chip is Anthropic discussing? Microsoft said in January that its Maia 200 processor was built for AI inference, the step where trained models generate answers and other outputs. Scott Guthrie, Microsoft’s executive vice president for Cloud + AI, wrote on January 26 that Maia 200 was designed to improve the economics of token generation. CNBC reported that Microsoft has not yet made the chip broadly available through Azure, even though the company said Maia 200 would run OpenAI’s GPT-5.2 model. (cnbc.com) Satya Nadella said on Microsoft’s April earnings call that Maia 200 offered “over 30% improved tokens per dollar” versus the latest silicon in Microsoft’s fleet, according to CNBC’s account of the call. CNBC also reported that the chips were already running in Microsoft data centers in Arizona and Iowa. ### How large is the Microsoft-Anthropic relationship already? Anthropic said on November 18, 2025 that Microsoft and Nvidia were entering broader strategic partnerships with the company. (blogs.microsoft.com) Anthropic said Microsoft had committed to invest up to $5 billion, while Anthropic committed to purchase $30 billion of Azure compute capacity and to contract additional compute capacity up to one gigawatt. Anthropic also said Claude models would be available through Microsoft Foundry and remain integrated across Microsoft’s Copilot products. (cnbc.com) CNBC reported on May 21 that Anthropic also relies on cloud services from Amazon and Google. Anthropic said in April that Amazon was investing another $5 billion, with up to an additional $20 billion over time, building on $8 billion previously invested. Anthropic said that agreement included up to 5 gigawatts of new AWS data-center capacity. ### Why is Anthropic still hunting for more compute? (anthropic.com) Dario Amodei said earlier in May that Anthropic had experienced “difficulties with compute,” CNBC reported. CNBC said the company’s Claude assistant and Claude Code programming tool had grown more popular this year, increasing the need for computing capacity. The same report said SpaceX disclosed Anthropic would pay $1.25 billion per month through May 2029 for computing power. (cnbc.com) Anthropic said in April that it would use Amazon Web Services’ Trainium chips in a 10-year arrangement worth more than $100 billion, according to CNBC. CNBC also reported that Anthropic announced plans in October to use Google tensor processing units. The result is a supplier list that already includes Amazon, Google, Nvidia and, potentially, Microsoft. ### How much money are private AI companies trying to absorb? (cnbc.com) 20minutos reported on May 23 that OpenAI, SpaceX and Anthropic could seek close to $200 billion from markets within a few quarters. The Spanish publication said SpaceX was aiming to raise between $75 billion and $80 billion, and described the three companies as attempting to absorb an amount of liquidity without recent precedent among private companies. That report framed the funding race as a test of how much capital public and private investors will continue to commit to a narrow set of AI-linked names. (cnbc.com) Bloomberg reported on April 20 that Amazon’s additional $5 billion investment in Anthropic valued the startup at $350 billion, excluding the new funding. Bloomberg also reported that Anthropic had completed a $30 billion funding round in February. ### Where does OpenAI’s governance fight fit into this? Elon Musk filed Musk v. Altman in federal court in Northern California on August 5, 2024, naming Sam Altman and multiple OpenAI entities as defendants, according to CourtListener’s docket. (20minutos.es) CourtListener shows the case remained active as of May 21, 2026, with the latest known filing dated May 20. Separate coverage in March 2025 reported that a judge denied Musk’s request to block OpenAI’s planned for-profit transition, leaving the broader case to continue. (bloomberg.com) The next public markers are likely to come from company announcements and court filings. Anthropic’s newsroom continues to post partnership updates, Microsoft has yet to announce external Azure availability for Maia 200, and the Musk v. Altman docket remains open in the Northern District of California. (anthropic.com) (courtlistener.com)