Musk Just Rented His Supercomputer to His Biggest AI Rival
Grab a coffee. This one's a movie.
Three days ago, Anthropic, the makers of Claude and Elon Musk's direct AI competitor, quietly signed a deal to take over the entire Colossus 1 data center in Memphis. The same supercomputer Musk built in 122 days to train Grok. The same one he called the most powerful AI cluster on Earth.
Now Anthropic owns every cycle of it.
We're talking over 222,000 Nvidia GPUs, including powerful H100 and H200 chips alongside next-generation GB200 accelerator systems, and 300 megawatts plus of compute power. To put that in perspective: 300 MW is enough electricity to power roughly a quarter of a million homes. Anthropic is plugging it all into making Claude faster. Tom's Hardware
The plot twist nobody saw coming
Here's where it gets wild. Musk and Anthropic's CEO Dario Amodei have been publicly trading shots for years over AI safety. Anthropic literally exists because its founders thought the AI race was moving too recklessly. Musk has called them "doomers."
And yet, Musk personally signed off. "Everyone I met was highly competent and cared a great deal about doing the right thing. No one set off my evil detector. So long as they engage in critical self-examination, Claude will probably be good", he posted on X. He also reserved the right to yank the compute back if Claude "engages in actions that harm humanity." Yahoo Finance
Yes, there's a kill-switch clause in the contract. Musk can unplug his rival mid-training run if he decides their AI has gone bad. That's not a business deal, that's a thriller script.
The reason he's comfortable doing this? "SpaceXAI had already moved training to Colossus 2". Translation: Colossus 1 is yesterday's news to him. He's already on the next thing. Yahoo Finance
The part that should rewire how you think about the future
Buried in the press release is this casual line: Anthropic "expressed interest" in working with SpaceX to develop multiple gigawatts of compute capacity in space. CNBC
Data centers. In orbit.
Why? Because we're hitting a wall on Earth. According to SpaceX itself, "The compute required to train and operate the next generation of these systems is outpacing what terrestrial power, land, and cooling can deliver on the timelines that matter". Cooling is free in space. Solar power is unlimited. Land doesn't exist as a constraint. Tom's Hardware
Sam Altman called the idea ridiculous a few months ago. Anthropic and SpaceX just put it on the roadmap.
What this actually means
For users of Claude, the immediate effect is concrete: Claude Code's five-hour rate limits have doubled for all paid tiers. The company has also removed the peak hours limit reduction for Pro and Max. If you've been hitting the wall, the wall just moved. Tom's Hardware
But the bigger story is the shape of the AI industry in 2026. Anthropic is now stitched into compute deals with Amazon (up to 5 GW), Google and Broadcom (5 GW), Microsoft and NVIDIA ($30 billion of Azure capacity), and a $50 billion U.S. infrastructure investment with Fluidstack. Add SpaceX. Add, eventually, space itself. Anthropic
The companies racing to build superintelligence have stopped pretending they're tech companies. They're energy companies, real estate companies, and now space companies. The new economy is being poured in concrete, copper, and silicon, and apparently, soon, in orbit.
Musk renting his supercomputer to the rival he keeps trash-talking isn't a contradiction. It's a confession: the demand for AI compute has gotten so insane that even mortal enemies are forced to do business. There's no "winning" the race anymore. There's only feeding it.
