Anthropic signs roughly $45 billion compute rental deal with UK startup Nscale

Anthropic commits roughly $45 billion to Nscale compute

Anthropic has signed a deal to rent approximately $45 billion worth of AI compute from Nscale, a British AI infrastructure company founded in 2024, according to a source familiar with the arrangement who spoke to TechCrunch. Bloomberg was first to report the agreement.

According to Bloomberg, the contract runs for six years, with the capacity delivered from Nscale's flagship data center in West Virginia. The compute will run on Nvidia's Vera Rubin platform, the chipmaker's newest flagship system, which combines six distinct chips designed to operate together and is regarded as the current frontier of chip design. TechCrunch's source said the capacity is expected to begin powering Anthropic's services in late 2027.

Nscale is a young company by any measure. It was founded only in 2024 and has already signed agreements with customers including Microsoft. Landing a contract of this magnitude from one of the best-funded AI labs would rank it among the largest AI infrastructure commitments made to date.

The latest move in an eight-month spending spree

The Nscale arrangement continues a run of compute deals Anthropic has struck over roughly the past eight months as it scales up capacity to compete with rivals, most prominently OpenAI.

Earlier in August, Anthropic signed a $10 billion agreement with Volta, an AI cloud startup founded earlier this year, securing a six-year supply of computing power from a data center in Norway. In July, it signed a $5 billion compute-related deal with chipmaker AMD.

In May, the company disclosed a large computing arrangement with SpaceX, the company run by Elon Musk, drawing capacity from two SpaceX data centers and reportedly worth $1.25 billion per month. In April, Anthropic significantly expanded its partnership with Amazon, gaining access to an additional 5 gigawatts of compute, and also deepened its relationships with Google and Broadcom.

A forward bet on next-generation silicon

The choice of hardware is as significant as the price tag. Because the capacity is not expected to come online until late 2027, the deal amounts to a forward purchase of compute on silicon that has yet to ship at scale. Anthropic is committing tens of billions today for capacity that will serve models and products several years from now, a sign of how far in advance AI labs must now secure chips, power and data center space to stay competitive.

It also illustrates the unusual shape of the AI supply chain that has emerged. Contracts once reserved for established hyperscale cloud providers are now flowing to very young specialists: Nscale was founded just two years before signing one of the largest compute deals on record, and Volta was only months old when it landed its $10 billion agreement.

Why it matters

A roughly $45 billion rental agreement is among the biggest AI infrastructure commitments yet, and it underscores that access to compute has become the central currency of the AI race. Anthropic is locking up capacity years in advance across every layer of the stack: hyperscalers like Amazon and Google, chipmakers like AMD, Nvidia's hardware ecosystem, and infrastructure startups barely out of the gate. TechCrunch notes that rivals including Google, OpenAI and Meta are pursuing similar strategies, so pressure on chip supply, electricity and data center capacity will keep intensifying. For infrastructure providers, the demand has inverted the usual startup dynamic: firms like Nscale and Volta are securing multi-billion-dollar contracts before they have had time to build a long track record, while AI labs take on enormous long-term financial commitments to guarantee they will not run short of computing power.