Nvidia, Microsoft Pour $15 Billion Into Anthropic for New AI Alliance
Anthropic also commits to purchase $30 billion from Microsoft cloud computing business Azure
- Nvidia and Microsoft will invest up to $15 billion in Anthropic, which will purchase $30 billion of compute capacity from Microsoft Azure.
- Anthropic will use advanced AI chips from Nvidia.
- The partnership highlights increasing collaboration among major AI players despite rivalries.
Three of the biggest companies in artificial intelligence announced a partnership featuring tens of billions of dollars in spending, adding to an investment spree that is aimed at supercharging AI model development.
Under the partnership, Nvidia NVDA -1.29%decrease; red down pointing triangle and Microsoft MSFT -2.47%decrease; red down pointing triangle will invest up to $15 billion in Anthropic, a competitor to OpenAI whose models are popular with coders and businesses. Anthropic, in turn, said it would buy $30 billion of computing capacity from Microsoft Azure running on Nvidia AI systems. The chip titan will also work with Anthropic on design and engineering.
The alliance is the latest sign in the global AI race that the biggest players must work with one another despite potential business rivalries and policy disagreements. Microsoft is a big investor in Anthropic rival OpenAI and offers its models to customers, while Nvidia Chief Executive Jensen Huang and Anthropic CEO Dario Amodei have clashed over chip exports and the number of jobs lost to AI.
The companies don’t want to lose out to competitors who are also plowing money into data centers for training models and offering new products.
“We are increasingly going to be customers of each other,” Microsoft CEO Satya Nadella said in a video statement with Amodei and Huang. Azure is deploying hundreds of thousands of Nvidia graphic processing units for training models.
Anthropic models will now be available on Azure, Microsoft said. The company already uses AI models from Anthropic in its 365 tools. It is considered critical for model developers to have access to customers of top cloud providers Microsoft, Amazon Web Services and Google to increase usage.
The new partnership is the latest in an array of deals between technology giants and hot AI startups aimed at stoking growth in the fast-evolving sector. Also Tuesday, Google launched Gemini 3, an updated version of its own model.
The flurry of deals and vast sums that technology titans and investors alike are plowing into new infrastructure have raised concerns in recent weeks about whether AI-powered products will generate enough revenue to justify it.
Nvidia and competitor AMD have both struck deals that have bolstered Anthropic rival OpenAI using funding from their own balance sheets, raising concerns among investors that the AI boom is being propped up by risky “circular” financing.
Anthropic has committed to initially buying up to 1 gigawatt worth of computing power from Microsoft—using the equivalent amount of electricity required to power roughly 750,000 homes for a year—that will be deployed using Nvidia’s Grace Blackwell servers, which employ the company’s current most-advanced AI chips, and its Vera Rubin systems, which are set to launch next year.
If the $30 billion doesn’t get to a gigawatt of capacity, Anthropic could spend more to get to that point, meaning the spending total would be even higher.
Concerns about an AI bubble have contributed to a recent dip in the stock market. Microsoft shares were down about 4% in Tuesday morning trading, while Nvidia shares were down about 3%, part of broad declines in the market.
Nadella said the partnership builds on Microsoft’s roughly $135 billion stake in OpenAI by providing Microsoft customers with more choice.
Microsoft and OpenAI recently struck a new deal that gives Microsoft a 27% stake in the startup’s new public-benefit corporation. The tech giant has broken off from the startup in some ways. Microsoft is pursuing its own AI breakthroughs through its Microsoft AI division, which plans to develop AI self-sufficiency from OpenAI by building, testing and releasing its own voice, image and text models.
Microsoft has access to OpenAI’s models until 2032, a schedule company executives have said gives the company time to make its own models into leading technology.
Microsoft AI CEO Mustafa Suleyman has praised OpenAI and the work the companies have done together, but he has also distanced himself from OpenAI and competitors in some ways, such as by declining to develop AI that simulates erotica.
In a recent interview with business streaming show TBPN, Nadella said the company is focused on being a “platform company” with many partners, and that it has its own capabilities of creating world-class AI technology. Numerous companies use Microsoft’s Azure cloud computing services to train their AI models and run workloads.
Amodei said Microsoft’s emphasis on enterprise platforms was a natural fit for Claude, which is frequently used in business applications.
Anthropic recently raised $13 billion from investors at a $183 billion valuation and has forecast turning a profit far more quickly than OpenAI, which plans to spend heavily on computing in the years ahead.
Anthropic is unique among AI companies in that it isn’t heavily reliant on Nvidia chips because it works closely with Amazon and Google.
A former OpenAI executive, Amodei has criticized Huang and the Trump administration for being willing to export AI chips to China and the Middle East, arguing that they are threatening national security and could boost China in the AI race.
Huang has said Amodei is being overly restrictive with his policy positions and pushed back earlier this year when the Anthropic CEO predicted massive white-collar job losses caused by AI.
“I pretty much disagree with almost everything he says,” Huang said at the time. He added that Amodei seems to think AI is so scary that only Anthropic should be building models and suggested it was a strategy to control the whole industry.
Amodei called that statement an “outrageous lie.”
The two executives complimented each other on Tuesday, with Amodei crediting Nvidia for fueling much of the AI boom and Huang saying Nvidia’s developers love Anthropic’s Claude models.
“Now your business is on a rocket ship,” Huang told Amodei.