Anthropic Volta US$10B deal secures key cloud computing capacity allocation

Anthropic has agreed to a six-year, $10 billion cloud-compute deal with Volta, a newly founded AI infrastructure startup, according to a Bloomberg report published August 4, 2026. The deal will see Volta partner with crypto-miner-turned-data-center-builder Bitdeer to build a 133-megawatt facility in Norway, running on Nvidia's next-generation Vera Rubin chip architecture. It follows a string of other major compute agreements Anthropic has signed in recent months.

Frontier AI labs are locking in compute capacity years in advance because GPU supply is tight and demand keeps rising. A $10 billion, six-year commitment from Anthropic to a company founded earlier this year demonstrates how much leverage new, well-funded "neocloud" providers now hold in the AI infrastructure market. Even startups with no long operating history can land deals this size if they can secure chips and build capacity fast enough.

Volta is part of Nvidia's Cloud Partner program, a group of AI cloud providers building their data centers around Nvidia GPUs. Before this report, Volta had publicly mentioned it was working with an AI lab on a deal but had not named which company. Bloomberg's report, citing anonymous sources, confirmed Anthropic as the customer.

The Norway facility will run on Nvidia's Vera Rubin systems, the chipmaker's newest AI chip architecture, which launched earlier in 2026. A 133-megawatt data center represents a substantial build for a company this new, and the partnership with Bitdeer—a firm that built its data center construction experience through crypto mining—reflects a broader pattern of mining infrastructure companies pivoting into AI compute as demand for GPU-heavy data centers has outpaced supply.

Anthropic has been expanding its compute capacity aggressively over the past several months as it competes with other frontier labs for training and inference capacity. The $10 billion Volta deal is one of several recent major commitments. Given the six-year term, capacity is likely to come online in phases rather than all at once, though timing has not been officially confirmed.

The deal reinforces a trend playing out across the AI industry: compute capacity, not just model quality, has become one of the biggest competitive battlegrounds for frontier labs. Companies willing to build fast, even those with no track record, can win contracts worth billions. Spreading infrastructure across multiple regions like Norway allows these massive energy demands to be met without hitting capacity walls at traditional grid infrastructure. Clean power hubs are becoming critical because scaling AI requires enormous electricity supply that traditional grids cannot handle alone. Distributing infrastructure across multiple providers reduces bottlenecks when companies stop relying on a single provider.