Anthropic signs computing deal with SpaceX to meet AI demand

[SAN FRANCISCO] Anthropic PBC has signed an agreement with Elon Musk’s SpaceX to bolster computing resources and meet surging demand for its Claude artificial intelligence software.

Anthropic plans to access more than 300 megawatts of computing capacity from a large SpaceX data centre in Memphis known as Colossus 1, the companies said on Wednesday (May 6). Anthropic said that the partnership would “substantially” increase its computing resources and enable it to raise usage limits for its AI products.

Terms of the deal were not disclosed. Other data centre developers have recently been able to lock in tenants at terms of anywhere from US$1.5 million to US$2 million per megawatt per year. At that rate, the Anthropic deal would easily represent hundreds of millions of US dollars worth of business for xAI annually.

The arrangement brings together two competitors in the race to develop more advanced AI systems. Anthropic has seen demand surge in recent months as more customers flock to use its tools to streamline the process of coding and other tasks. In addition to SpaceX, Anthropic has also tapped Alphabet’s Google for chips and cloud services.

At a conference after the SpaceX deal was announced, Anthropic CEO Dario Amodei said that his company is “working as quickly as possible” to secure more computing resources after experiencing “80x growth” in annualised revenue and usage in the first quarter of this year.

Musk, a fierce critic of Anthropic rival OpenAI, said that he spent time last week with senior staffers at Anthropic and decided to lease Colossus 1 to the company after making sure “Claude is good for humanity”.

“No one set off my evil detector,” he said. “So long as they engage in critical self-examination, Claude will probably be good.”

Still, Anthropic is a competitor to xAI, including for talent. Ross Nordeen, one of xAI’s founding members, said on Wednesday that he was joining Anthropic’s compute team. Nordeen had spent years at Tesla before moving to xAI, and left the startup about a month ago. He was the last xAI cofounder to depart the company, where he also focused on computing power. Musk is the only founder left at xAI.

Though Musk’s xAI is behind on coding, the AI startup, now part of SpaceX, has been ahead of some in building data centres and stockpiling Nvidia chips. XAI’s computing deals could position the company as an infrastructure provider and drive up revenue as SpaceX prepares to go public.

SEE ALSO

Anthropic unveils AI agents to field financial services tasks

Anthropic commits to spending US$200 billion on Google’s cloud and chips: report

Anthropic considering funding offers at over US$900 billion value

Anthropic, OpenAI splurge on London office space in leasing wave

More

Musk’s company previously agreed to provide some of its computing resources to AI coding startup Cursor as part of a deal that would see both firms collaborate. XAI has been building data centres in Tennessee and Mississippi, and has raised capital to rent the chips for the facilities.

On Wednesday, Musk said that xAI would be “dissolved as a separate company”. Going forward, he said. “It will just be SpaceXAI, the AI products from SpaceX.”

Musk has been reorganising xAI by bringing in personnel from SpaceX to help oversee the firm, lead software infrastructure and optimise chips usage. XAI’s compute utilisation rate for its chips was running at only 11 per cent, according to an internal memo last month, suggesting the company had more computing resources than needed solely for its own AI efforts.

As part of the new agreement, Anthropic said that it has also “expressed interest in partnering with SpaceX to develop multiple gigawatts of orbital AI compute capacity”. Musk has previously said that he wants to put data centres in space and tap into solar power. BLOOMBERG