Anthropic signs $45 billion Nscale deal to lock in AI compute

Anthropic continues its compute-gobbling streak with a deal to rent roughly $45 billion in AI compute from Nscale, a British infrastructure company founded in 2024, a source familiar with the agreement told TechCrunch. The six-year pact will supply Nvidia Vera Rubin chip capacity from Nscale's flagship West Virginia data center, with power expected online in late 2027.

The agreement, first reported by Bloomberg, is the latest move in an aggressive expansion reshaping how frontier AI labs secure the horsepower behind their models. For Anthropic, more compute means faster iteration, stronger models, and a better shot at keeping pace with rivals in a market where capacity has become as strategic as algorithms.

Key Takeaways

• Anthropic will rent about $45 billion in AI compute from Nscale over six years.
• Capacity will run on Nvidia's Vera Rubin chips from a West Virginia data center.
• Compute is expected to power Anthropic services starting in late 2027.
• The deal follows a rapid string of partnerships with Volta, AMD, SpaceX, Amazon, and Google.
• Google, OpenAI, and Meta are pursuing similar compute hoarding strategies.

Why is Anthropic spending billions on Nscale compute?

Over the past eight months, Anthropic has aggressively scaled its compute capacity to better compete with rivals, most notably OpenAI. The Nscale deal adds another massive block of rented infrastructure without requiring Anthropic to build and operate every data center itself.

Nscale, despite being founded only in 2024, has already secured deals with major players including Microsoft. That track record likely made the British provider an attractive partner for a lab racing to lock in long-term capacity before demand outstrips supply.

What is Nvidia's Vera Rubin and why does it matter?

Under the agreement, Nscale will supply Anthropic with compute powered by Nvidia's Vera Rubin chips, described as a new state-of-the-art chip system. The Vera Rubin platform combines six different chips working in concert and is considered the cutting edge of chip design.

For Anthropic, access to next-generation hardware arriving in late 2027 could help sustain model improvements well into the next product cycle. In the broader Future Tech & AI Wonders landscape, chip architecture has become a front-line battleground alongside model training itself.

How does this deal fit Anthropic's broader compute strategy?

The Nscale pact is only the latest entry on a crowded ledger. Earlier in August, Anthropic signed a $10 billion deal with Volta, an AI cloud startup founded in January, securing six years of cloud computing from a data center in Norway. In July, it also closed a $5 billion compute-related deal with AMD.

May brought a large SpaceX arrangement drawing capacity from two data centers, reportedly worth $1.25 billion per month. In April, Anthropic expanded partnerships with Amazon for an additional 5 gigawatts of compute and broadened ties with Google and Broadcom for more power capacity.

Who else is racing to hoard AI compute?

Anthropic is hardly alone. The industry-wide race to gobble up compute capacity continues, with Google, OpenAI, and Meta all following a similar path. OpenAI's primary advantage over rivals has long been its investment in compute, though recent executive turnover—including the departure of head of data centers Chris Malone—has raised questions about internal stability even as demand for capacity grows.

Both Anthropic and OpenAI are also reportedly planning public debuts, making long-term infrastructure commitments like the Nscale deal part of a wider financial chess match. As frontier labs stack multi-billion-dollar agreements, the AI arms race is increasingly measured in gigawatts, not just parameters.