Anthropic Buys $10B of Compute From Six-Month-Old Volta

Would you sign a $10 billion contract with a company that didn't exist six months ago? Anthropic just did. The maker of Claude has locked in ten billion dollars of computing capacity over six years from Volta Infra Holdings, a cloud startup founded in early 2026 by former Brookfield executives. The capacity arrives in two phases through March 2027.

A bitcoin mine, Norwegian hydropower, and Nvidia's newest chips

Volta's short history reads like a compressed summary of this AI cycle. Months after founding, it raised $300 million led by Andreessen Horowitz and Altimeter Capital, with Nvidia and Michael Dell among the backers, at a $2.4 billion valuation. The capacity it will serve Anthropic sits in Tydal, Norway: a 133-megawatt, hydropower-fed data center stocked with Nvidia's new Vera Rubin chips. The facility's operator is Bitdeer Technologies, a bitcoin miner, one more crypto operation pivoting its megawatts toward AI hosting. Volta has also assembled a $5 billion financing pool to help customers carry the cost of the chips themselves, so this is as much a financing vehicle as it is a cloud provider. Attentive readers will notice the loop: Nvidia is an investor in a company whose business is buying Nvidia chips. Critics have a name for this pattern, circular financing, and this deal is about as clean an example as the industry has produced.

From Anthropic's side, the deal extends a deliberate strategy of spreading compute across suppliers: Google's TPUs, Amazon's data centers, the recently announced $5 billion AMD agreement, and now Volta. At this scale, dependence on a single provider is a risk nobody is willing to carry anymore.

Google quietly restructured its own exposure the same week

The Volta deal landed alongside a related story: Google has moved billions of dollars in chip commitments it carries for Anthropic off its own balance sheet, through a financing structure involving Broadcom, Apollo, Blackstone, and Morgan Stanley. The largest AI infrastructure bets are increasingly executed not as straightforward purchases but through special-purpose financing, leases, and third-party capital.

Reading the two deals together

Taken together, these stories say the cost of AI infrastructure has outgrown what even the wealthiest players want to carry alone, and the industry is responding with progressively more creative financial engineering. There are two honest readings. The optimistic one: capacity is scaling fast, competition among providers is intensifying, and some of that eventually shows up as lower prices for everyone building on these models. The cautious one: ten-billion-dollar commitments to six-month-old companies, backed by off-balance-sheet structures, concentrate fragility even as they spread cost. For a business buying AI services, the practical takeaway is to enjoy the falling prices this arms race produces, while being deliberate about how deeply you couple your operations to any single provider whose own foundations are still being financed in real time.

Sources: The Decoder (Volta), The Decoder (Google)