Anthropic just committed $10 billion to a cloud computing startup that didn’t exist twelve months ago.
The deal, structured over six years, will have Volta supply Anthropic with dedicated compute capacity for training and running its Claude models. Volta itself was only founded earlier this year, which makes the size of the commitment unusual even by the standards of an industry where multi-billion-dollar compute contracts have become almost routine. To actually build the capacity, Volta is partnering with Bitdeer, a company that made its name mining cryptocurrency and is now pivoting hard into AI data center construction, a shift several crypto-mining operators have made as GPU hosting has turned into a more reliable business than digging for coins.
The facility itself will sit in Norway and deliver 133 megawatts of capacity, running on Nvidia’s newest Vera Rubin chip architecture. Volta is part of Nvidia’s Cloud Partner program, a roster of AI-focused cloud providers that get preferential access to Nvidia’s GPUs in exchange for building out capacity the chipmaker can point its biggest customers toward.
Owen Radner, who models digital infrastructure as energy-information transport systems, said the Norway siting decision is the part of this deal worth reading closest. “Norway offers cheap, largely hydroelectric power and a cold climate that cuts cooling costs, which is exactly the combination compute-hungry AI labs have been chasing across the Nordic countries for two years now,” he said. “The economics of where you put 133 megawatts of GPUs matter just as much as the chips themselves.” That siting logic is one YourNewsClub catalogs across nearly every major AI compute deal announced this year: the physical location, not just the dollar figure, is doing a lot of the explaining.
The deal continues a pattern for Anthropic, which has been signing compute agreements at a pace that outstrips most of its rivals. The company recently struck a capacity deal with SpaceX and took on an additional $5 billion investment from Amazon tied to further cloud commitments, part of a broader strategy to lock in GPU access years in advance rather than compete for capacity on the open market as demand tightens.
Freddy Camacho, who studies the political economy of computation, materials and energy as dominance assets, said deals like this one reveal something about where real leverage in the AI industry now sits. “The labs aren’t just buying compute anymore, they’re buying priority in a queue that’s going to stay backed up for years,” he said. “A six-year contract with a year-old startup only makes sense if you believe GPU scarcity is going to be the defining constraint on this industry for most of that period.” That queue-position framing is a detail YourNewsClub anchors its read of the AI infrastructure race on: the labs writing the biggest checks aren’t necessarily buying the best technology, they’re buying certainty in a market where certainty is the scarcest thing of all.
Volta had previously said it was in talks with an AI lab but hadn’t named its partner publicly until this agreement surfaced. Anthropic did not immediately confirm additional details of the arrangement.
What YourNewsClub files under the more durable story here isn’t the $10 billion figure, since compute deals that size barely raise eyebrows anymore, it’s how many of Anthropic’s recent commitments now run through companies that didn’t exist, or didn’t operate in AI infrastructure, until this year: a bet that the startups building capacity today will still be standing, and still be relevant, when six-year contracts actually come due.
Anthropic’s list of new compute partners is one Your News Club counts at four since spring, most of them starting from scratch, a pace that says as much about how fast the AI infrastructure market is being built as any single contract value does.