Anthropic’s $45 Billion Bet Turns Compute Into the New AI Battleground

Anthropic’s $45 Billion Bet Turns Compute Into the New AI Battleground
Anthropic is making an enormous wager that access to computing power—not just better models—will decide the next phase of the AI race. Its latest agreement would hand a young British infrastructure company a central role in that contest.

The reported deal, first outlined as a $45 billion arrangement with UK start-up Nscale, gives Anthropic access to a vast pool of AI compute over six years. Nscale, founded in 2024, is expected to supply the capacity from its flagship West Virginia data centre, using Nvidia’s Vera Rubin system, a new high-end chip platform built around six chips working in concert.

The timetable is long, but the stakes are immediate: the capacity is expected to begin powering Anthropic’s services in late 2027. That means the company is committing years ahead, treating future computing supply as a strategic asset rather than a routine operating expense.

The Nscale agreement is also the latest step in a rapid expansion. In April, Anthropic broadened its Amazon partnership for an additional 5 gigawatts of compute and expanded work with Google and Broadcom. In May, it disclosed a major SpaceX arrangement reportedly worth $1.25 billion in capacity each month. A $5 billion AMD-related deal followed in July, and earlier this month Anthropic secured a six-year, $10 billion cloud-computing agreement with Volta in Norway.

The scale of the new deal stands out even within that buying spree. A source familiar with the arrangement described it as a plan to “rent about $45 billion in AI compute” from Nscale. Anthropic’s approach mirrors the wider industry’s scramble: OpenAI, Google and Meta are all pursuing more infrastructure, but Anthropic is tying its competitive ambitions to a chain of long-term capacity bets.

Continue reading https://foxvector.com/stories/01a046ae-233f-21bd-73d5-20de4e415080