Anthropic has signed a deal to lease AI computing from British AI infrastructure company Nscale for about $45 billion, people familiar with the deal told TechCrunch.
Although Nscale was only founded in 2024, it has already signed deals with Microsoft and others to supply computing to Anthropic through its new cutting-edge chip system, Nvidia’s Vera Rubin chips.
The Vera Rubin system combines six different chips that work together and is considered the most advanced chip design. The computing power will begin supplying AI Lab services in the second half of 2027, the person said.
The deal, first reported by Bloomberg, is just the latest in a series of computing partnerships for Anthropic. Bloomberg writes that the deal will span six years and that the computing power will be sourced from Nscale’s flagship data center in West Virginia.
Over the past eight months, Anthropic has been aggressively scaling up its computing power to better compete with its competitors, particularly OpenAI.
Earlier this month, Anthropic signed a $10 billion deal with Volta, an AI cloud startup founded in January, securing a six-year supply of cloud computing power from its data centers in Norway. The company also signed a $5 billion computing deal with AMD in July.
A few months earlier, in May, Anthropic announced a major computing deal with SpaceX (run by Elon Musk). Musk is unlikely to be an ally of Anthropic due to his conflict with OpenAI CEO Sam Altman. The deal reportedly draws computing power from two different SpaceX data centers and provides Anthropic with $1.25 billion worth of capacity per month.
In April, Anthropic also signed a deal to significantly expand its partnership with Amazon, gaining access to an additional 5 gigawatts of computing. That month, the company also expanded its relationships with Google and Broadcom, adding more power capacity.
Anthropic is not alone in its passionate pursuit of improving AI capabilities. Other major companies, including Google, OpenAI, and Meta, are following a similar trajectory, and the race to consume as much computing power as possible continues.
If you buy through links in our articles, we may earn a small commission. This does not affect editorial independence.
