2 min read AI-generated

Anthropic Locks In Compute: A $10 Billion Deal With Volta for a Norwegian Data Center

Copy article as Markdown

Anthropic has signed a six-year, $10 billion deal with Volta — a startup that's only seven months old. The compute comes from a hydro-powered data center in Norway, running Nvidia's Vera Rubin chips.

Featured image for "Anthropic Locks In Compute: A $10 Billion Deal With Volta for a Norwegian Data Center"

The numbers in AI keep getting bigger, and this deal is no exception: according to a Bloomberg report on August 4th, Anthropic has signed a six-year, $10 billion agreement for cloud compute. The partner is Volta, an AI infrastructure startup founded only in January 2026 — so barely seven months old.

What’s in the deal

Volta secures Anthropic roughly 121 megawatts of Nvidia Vera Rubin capacity. It’s being built at Bitdeer’s Tydal site in Norway — a Bitcoin miner that’s putting up a 133-megawatt hydro-powered data center there. Capacity is meant to arrive in two phases, targeting December 31, 2026 and March 31, 2027.

The division of labor is telling. Volta leases the chip capacity and helps clients finance these costly deals. Bitdeer builds and runs the hardware. And Anthropic is the customer, locking in compute years ahead of time. Norway isn’t a coincidence either: hydro power is cheap, steady, and fits the growing pressure not to run AI data centers entirely on fossil energy.

My take

Deals like this say more about the state of the industry than any benchmark. Frontier labs are now committing to individual chip deliveries years out because GPUs are scarce and demand won’t let up. That Anthropic is betting on a seven-month-old startup rather than one of the big cloud providers shows how creative the financing has become — and how tight the market is.

Two things I’m taking away. First: the real bottleneck in AI is no longer the model, but the power and the chips behind it. Whoever locks in both early buys themselves time. Second, this is another piece in the debate about AI debt and concentration risk. Ten billion over six years is a bet that demand stays. I find the bet reasonable — but it also shows how much capital is riding on very few assumptions right now.


Sources: