Just a few months ago, Elon Musk accused the AI company Anthropic of stealing artificial intelligence training data "at massive scale" in a post on his social network X.
That apparently hasn't stopped the billionaire from doing business with the company. Musk's SpaceX has signed a data center deal that will give Anthropic access to more than 200,000 Nvidia GPUs worth of power at its Colossus 1 supercomputer facility in Tennessee.
The partnership will give Anthropic additional firepower to "directly improve capacity for Claude Pro and Claude Max subscribers," SpaceX said in a website post. "As part of this agreement, Anthropic also expressed interest in partnering to develop multiple gigawatts of orbital AI compute capacity."
Because of this deal, Anthropic said in its own post, the company is raising usage limits for users across some of its products. The changes, effective immediately, double Claude Code rate limits for users of Claude on Pro, Max, Team and seat-based Enterprise plans, remove peak-hour restrictions of Claude Code for Pro and Max accounts and raise API limits for Claude Opus models.
Watch this: Watch SpaceX Starship Flight Test 12 End With a Bang
More AI means more data center deals
In the same post, Anthropic listed some of its other data center agreements with companies, including Amazon, Google and Microsoft, and reiterated its intention to keep expanding internationally. In the era of data center backlashes, Anthropic also announced in February that it has pledged to cover the costs of energy price increases driven by data center activity. Critics have questioned how companies such as Anthropic can uphold those pledges.
The deal with SpaceX, which acquired Musk's AI company xAI earlier this year, may have surprised some, but AI companies are scrambling to secure data center resources as they continue to develop increasingly data-hungry artificial intelligence models.
At the same time, some communities are pushing back on new data center construction, leading some in the industry, Musk in particular, to plan to build data centers in space.
Among the groups criticizing the deal is the NAACP, which said in a statement about SpaceX, "Any company that disregards the obvious environmental and health concerns of Black communities to supposedly power a future that will help us all is sending a clear message about who it intends to serve in that future... Anthropic's use of a data center that pollutes a historically Black community is, at best, an uninformed decision, and at worst, a total disregard for the community's wishes and health."
The organization pointed to a lawsuit it has filed against SpaceX over environmental concerns at its Colossus 1 computing center.
Infrastructure for AI companies is 'mission critical'
It's never easy to gauge what Elon Musk will do next, but his decision to align with Anthropic appears to be a pragmatic one as the major AI giants align partnerships and as SpaceX tries to scale up its data business quickly ahead of a summer initial public offering. Public beefs may not matter when billions are at stake, and Anthropic may need SpaceX's hardware as much as SpaceX needs Anthropic's business.
"Securing data center resources is becoming mission critical for all major AI players," Lu Zhang, founder and managing partner at Fusion Fund, told CNET. "The bottleneck is no longer just model capability, but access to compute, energy, and physical infrastructure."
SpaceX, Zhang said, is unique among companies offering data center service because it can access large amounts of capital, has established relationships in the energy industry, and moves quickly, in addition to having its own AI business.
Zhang, who focuses on enterprise AI business trends, said that major AI companies risk deployment constraints and higher costs if they don't secure their long-term infrastructure now.
"Over time, compute access may become as strategically important as the models themselves," she said.
