Anthropic signs $45bn AI compute agreement with UK infrastructure firm Nscale
The AI laboratory has committed $45 billion to a British infrastructure startup for next-generation computing power, highlighting the escalating global race to secure hardware dominance.
Anthropic has agreed to rent approximately $45 billion worth of artificial intelligence computing power from Nscale, a British AI infrastructure company. The six-year agreement will draw capacity from Nscale’s flagship data centre in West Virginia.
The supply will be powered by Nvidia’s new Vera Rubin chip system, which integrates six different chips to deliver state-of-the-art performance. This computing capacity is scheduled to begin supporting the AI laboratory’s services in late 2027.
Nscale, despite being founded only in 2024, has rapidly established itself in the market, securing agreements with major technology firms including Microsoft. This partnership underscores how European-born infrastructure providers are becoming critical nodes in the global AI supply chain, even when the physical data centres are located in the United States.
The Nscale agreement is the latest in a rapid sequence of infrastructure investments by Anthropic over the past eight months. The company is aggressively scaling its hardware access to close the competitive gap with rivals such as OpenAI.
Earlier this month, Anthropic committed $10 billion to Volta, an AI cloud startup founded in January, for a six-year supply of computing power from a facility in Norway. In July, the company finalised a $5 billion compute-related arrangement with AMD.
Prior to that, Anthropic struck a major deal in May with SpaceX to draw capacity from two of its data centres, reportedly valued at $1.25 billion per month. The company also expanded its existing partnerships in April, gaining an additional five gigawatts of compute from Amazon while deepening ties with Google and Broadcom.
This relentless accumulation of processing power reflects a broader industry imperative. Major technology players, including Google, OpenAI and Meta, are engaged in a capital-intensive race to monopolise advanced semiconductor capacity before it becomes available.
For investors and European policymakers, these multi-billion-dollar commitments signal that AI infrastructure is now the primary battleground for market share. The reliance on next-generation hardware will dictate which laboratories can deploy the most capable models over the coming decade.
