Three months ago Elon Musk wrote to x That Anthropic is “evil,” “an anti-human,” and that the AI laboratory hates Western civilization. On Wednesday, it leased one of its most valuable assets, Anthropic, the world’s largest supercomputer.
But anthropic lovers shouldn’t rejoice too much in Musk’s newfound praise (even though he may have decided “no one set off my evil detector”). Analysts said the deal had little to do with them as a company Luck, and everything related to an upcoming brochure.
SpaceX It is expected to begin its public roadshow next month a confidential S-1 has been filed A valuation between 1.75 trillion and 2 trillion dollars is targeted for April 1. Wednesday’s announcement, along with Musk spun off his AI company xAI to SpaceX (to make SpaceXAi), gives the IPO something it didn’t have a week ago: a major AI customer for a reliable cloud infrastructure business.
The Anthropic deal will provide SpaceX with annual revenues of $3 billion to $4 billion and cash profits of more than $2.5 billion, according to estimates by New Street Research analyst Antoine Chkaiban. The margins seem excessive, but that’s because the data center is already built: the fixed capex is sunk, and the only meaningful operating costs are electricity plus relatively minimal staff costs.
“He won’t want multi-billion dollar GPUs to sit idle,” Chkaiban said. Luck. “This is a very good business decision.”
And it appears to be the beginning of Elon Musk’s transition from seeking to be a pioneer in the model race to becoming the owner of artificial intelligence.
“Whoever controls the data center now really controls the implementation of artificial intelligence,” said former president Andrew Moore. Google CEO of Cloud AI and now defense AI startup Lovelace AI, said: Luck. “So yeah, I think both parties to this wedding of convenience are going to be a little stressed out by it.”
Hyperscaler pivot
Colossus 1 contains roughly 220,000 Nvidia GPUs and was built in 2024 to train Grok, Musk’s AI assistant. But Grok didn’t fill it. Chkaiban estimates Grok’s annual revenue at less than $1 billion; Anthropic is on track to generate more than $40 billion in revenue. Inequality in agreement. Musk has too much computing power and Grok can’t fill it despite endless “ask Grok” questions about X; Anthropic has too many users and not enough processing power. The lease of Colossus 1 to Anthropic funds this gap.
But it also allows Musk to skip a step. The biggest cost item for any frontier AI lab is the 30%+ margin paid to AWS. Microsoft Azure or Google Cloud for computing. SpaceX is capturing the margins of hyperscalers rather than paying them off with stressful debt deals like AI labs.
Analysts said this framework (of SpaceX as the fourth hyperscaler) is what Musk wants investors to accept before pricing Luck. A SpaceX that can compete with AWS is worth a hyperscaler multiplier, not a rocket company multiplier. Nowadays Alphabet, Microsoft and Amazon trade for roughly double forward earnings Boeing’s And LockheedMartin.
But Moore doubted the pivot would be easy. large corporate customers such as governments or Luck 500 companies choose where to store their data centers mostly by location; if the cost of energy is cheap, if something goes wrong, they have fail-safe. Building a massive data center in Memphis wouldn’t replicate AWS’s global and regulatory footprint. “The war is not just about who has the most computing servers,” he said.
Moore added: “I would never disagree with Elon doing something great, but he has his work cut out for him to really beat AWS.”
Kill-switch sentence
Whether Musk wins this battle or not, he already has something other computing providers don’t. In his response to X, he wrote that SpaceX “reserves the right to take back the computer” if Anthropic’s AI “engages in actions that harm humanity.” The clause was not included in the official press release and it is unclear whether it was included in the contract. But if feasible, it would give Musk strong control over one of the three current leading AI labs while he sues OpenAI’s leadership in federal court.
This means Musk has more power than he did two weeks ago. And this would matter less if Musk hadn’t changed his mind about AI so much.
Moore, who was dean of computer science at Carnegie Mellon during Musk’s loudest phase of existential risk, remembers him as “one of the loudest voices saying that artificial intelligence was an existential threat to the human race.” Now he says artificial intelligence will lead to a world of abundance.
Antropik almost certainly has plans for a comeback. Frontier AI labs isn’t in the habit of single-sourcing the data center on which all its products depend, and Moore said the company will work aggressively on computing efficiency in the background. “There will be contingency plans in three months, six months, twelve months,” he said.
Still, neither side can break up cleanly. Gene Munster, managing partner of Deepwater Asset Management, puts the chances of the deal still being valid in two years at 80%. The remaining 20% is a bet on Musk himself. “Elon’s history is what makes it unique,” Munster said. “He might change his mind. It’s not about the actual procurement; it’s more about who is handling the procurement.”
The chances of the agreement are unquestionable. Even if Munster is right and the contract is valid for two years, one of the three leading AI labs in the world is now running on infrastructure controlled by a rival’s CEO.
“The risks are huge,” Moore said. “Everybody’s just trying to get through the next six months. They’ll do whatever it takes.”
This story first appeared on: Fortune.com
