Asger Risom
Elon Musk has struck a multibillion-dollar infrastructure deal with Anthropic just months after calling the AI company “evil,” as SpaceX positions itself to become a major hyperscale computing provider for the global AI industry.
Just months after publicly attacking Anthropic as “evil” and “misanthropic,” Elon Musk has struck a deal that could turn one of his companies into a critical infrastructure provider for the AI industry.
The agreement gives Anthropic access to Colossus 1, Musk’s massive AI supercomputer cluster, while potentially reshaping SpaceX into something much larger than a rocket company: a hyperscale AI infrastructure business competing with Amazon, Microsoft and Google.
A sudden alliance
The partnership comes after months of escalating rhetoric from Musk against rival AI firms.
In February, Musk wrote on X that Anthropic hated “Western civilization” and described the company as dangerous. This week, however, SpaceX announced a major compute leasing agreement with the AI startup, which analysts estimate could generate between $3 billion and $4 billion annually.
According to estimates cited by Fortune, much of that revenue could translate directly into profit because Colossus 1 has already been built. The largest ongoing operational costs are now power consumption and staffing rather than infrastructure construction.
The facility reportedly contains around 220,000 Nvidia GPUs and was originally developed to train Grok, Musk’s own AI assistant. But analysts believe Grok’s commercial growth has lagged far behind rivals.
While Grok is estimated to generate under $1 billion annually, Anthropic is reportedly on track for more than $40 billion in yearly revenue, creating a major imbalance between available computing capacity and demand.
That gap appears to be driving the agreement.
“He’s not going to want multiple billions of dollars of GPUs sitting idle,” New Street Research analyst Antoine Chkaiban told Fortune. “It’s a very good business decision.”
The real AI battle
The deal also signals a broader shift in Musk’s AI strategy.
Rather than competing solely as a model developer through Grok, SpaceX is increasingly positioning itself as an infrastructure provider selling computing power to AI companies desperate for capacity.
That would place SpaceX in direct competition with hyperscalers like Amazon Web Services, Microsoft Azure and Google Cloud, which currently dominate AI infrastructure markets.
For frontier AI labs, computing costs have become one of the industry’s biggest financial pressures. Training and running advanced AI systems requires enormous clusters of GPUs, consuming huge amounts of electricity and forcing companies into increasingly expensive cloud agreements.
By leasing out Colossus 1 directly, SpaceX can capture those infrastructure margins itself rather than paying them to outside providers.
Analysts told Fortune the timing is significant because SpaceX is reportedly preparing for a public roadshow ahead of a potential IPO valued between $1.75 trillion and $2 trillion.
Presenting SpaceX as a future AI infrastructure giant rather than purely a space company could dramatically increase how investors value the business.
“A SpaceX that can compete with AWS is worth a hyperscaler multiple, not a rocket company multiple,” the report noted.
One company controls the servers
The agreement also raises uncomfortable questions about power concentration inside the AI industry.
Following the announcement, Musk wrote on X that SpaceX “reserves the right to reclaim the compute” if Anthropic’s AI “engages in actions that harm humanity.”
It remains unclear whether that clause exists formally inside the contract. But if enforceable, it would give Musk extraordinary leverage over one of the world’s leading AI companies while simultaneously running his own competing AI operation.
Andrew Moore, former head of Google Cloud AI and current CEO of Lovelace AI, described the arrangement as unusually risky.
“He who controls the data center, really does control the application of artificial intelligence right now,” Moore told Fortune.
The situation is especially unusual because Musk’s views on AI have repeatedly shifted over time.
For years, Musk warned that artificial intelligence posed an existential threat to humanity. More recently, however, he has promoted AI as a transformative force capable of creating abundance and economic expansion.
Analysts believe Anthropic almost certainly has contingency plans and alternative infrastructure arrangements in development to avoid depending entirely on Musk-controlled systems.
Still, the partnership highlights how rapidly the AI industry is consolidating around access to compute power rather than simply software models.
As AI competition intensifies, ownership of the servers themselves may become just as strategically important as the technology running on them.
Sources: Fortune
