Anthropic Secures $10B AI Cloud Deal with Volta

What Anthropic Is Getting

Volta will supply cloud compute resources to Anthropic over the next six years through a partnership with Bitdeer, a crypto-mining company. The two firms are building data center infrastructure in Norway with a planned capacity of 133 megawatts. The facility will run on Nvidia’s latest AI chip architecture, the Vera Rubin systems.

Volta is also part of Nvidia’s Cloud Partner program, a consortium of AI cloud providers integrating Nvidia GPUs into their data centers.

Part of a Broader Compute Blitz

This deal is one piece of Anthropic’s aggressive push to secure computing power as it competes with rivals racing to scale AI models. The company has announced other significant compute partnerships recently, including agreements with:

• SpaceX
• Amazon

The pattern is clear: Anthropic needs more compute to train and refine Claude models while competing with OpenAI, Google, and other labs also racing to scale their systems.

Raw compute power has become the defining competitive advantage in AI. Access to vast GPU clusters separates leading labs from everyone else. For Anthropic, this deal ensures it can continue developing Claude at the pace the market demands without falling behind competitors who already have massive infrastructure advantages.

But the larger story here is that compute is becoming the bottleneck. Nvidia controls the chips. Cloud providers control the data centers. And companies like Anthropic are increasingly dependent on partnerships with startups like Volta just to get access to the resources they need. That’s a precarious position for any AI firm trying to remain independent.

The Open Question

The deal also raises a question worth asking: how much compute is actually enough? With $10 billion committed to six years of infrastructure, Anthropic is betting that scaling Claude’s training will continue to drive better models. That assumption has held true so far. But it’s not guaranteed that more compute always equals better outcomes, especially as models get larger and efficiency gains slow.

For now, though, Anthropic has secured the resources to keep running the race. Whether it’s running in the right direction is a separate question.