🤚 The Open-Palm Illumination

In what may be the most deliciously petty strategic pivot in Silicon Valley history, Elon Musk — the man who once called Anthropic’s Claude AI “misanthropic and evil” — has just leased the company 300 megawatts of computing capacity and access to over 220,000 Nvidia GPUs from xAI’s Colossus 1 supercomputer in Memphis, Tennessee.

Let that marinate. The man who sued OpenAI for betraying humanity’s interests is now personally funding OpenAI’s most dangerous competitor. This isn’t a business deal. This is a $852 billion grudge match being waged through data center leasing agreements, and honestly, we’ve never been more impressed by the genre.

The timing, naturally, is exquisite. This deal was announced on May 6th, 2026, smack in the middle of Week 2 of the Musk v. OpenAI trial, where Elon is attempting to get Sam Altman ousted from OpenAI’s board while arguing the company betrayed its nonprofit founding mission. And what better way to underline your legal argument that OpenAI has gone rogue than to hand its primary rival a continent-sized GPU cluster?

👐 The Two-Handed Reality Check

Let’s be clear about what’s actually happening here, because the layers of irony are structurally load-bearing.

xAI merged with SpaceX back in February 2026, making Musk’s AI company a subsidiary of his rocket company — because when your portfolio includes electric cars, brain implants, tunnel-boring machines, and social media platforms, why not consolidate the AI division under the space agency. It’s the obvious organizational chart.

With xAI’s Colossus 1 facility sitting on a mountain of underutilized compute, Musk had a decision: let the GPUs idle, or lease them to someone who would use them to compete directly with the company he’s suing for billions of dollars. He chose the latter. Capitalism doesn’t do irony — it does quarterly revenue.

Meanwhile, Anthropic was desperate. The company had been forced to ration customer usage due to a crushing shortage of computing infrastructure, falling behind OpenAI, which has been building massive data centers with SoftBank’s backing. After securing the SpaceX deal, Anthropic immediately doubled the usage limits for paid Claude Code users — a move that suggests they were running their inference clusters the way college students run their checking accounts: constantly, dangerously close to zero.

And then there’s the remarkable conversion story. Musk reportedly spent a week meeting with senior Anthropic leadership before approving the deal, after which he wrote: “Everyone I met was highly competent and cared a great deal about doing the right thing. No one set off my evil detector.” The evil detector, apparently, has a very specific calibration range, and it’s set somewhere between “people I’m currently suing” and “everyone else.”

🌿 The Gentle Awakening

Step back from the personal drama and the picture becomes almost philosophical. We have entered an era where the companies controlling large Nvidia clusters are starting to look just as important as the companies building the models themselves. Infrastructure is the new kingmaker. Compute is the new oil. And the landlords of the GPU datacenters are now more powerful than their tenants.

Musk understands this intuitively, which is why this deal accomplishes three things simultaneously:

• Revenue generation — turning idle compute into cash ahead of SpaceX’s expected IPO next month
• Strategic warfare — strengthening OpenAI’s most credible rival while suing OpenAI in federal court
• Narrative control — positioning himself as the reasonable benefactor of responsible AI development, in contrast to the company he says went rogue

The Musk v. OpenAI trial, meanwhile, continues to produce testimony that belongs more in a prestige HBO drama than a courtroom. Shivon Zilis — former OpenAI board member and mother of Musk’s children — testified that Musk once offered Altman a Tesla board seat. Musk admitted under oath that xAI distills OpenAI’s models. Settlement texts were sent two days before trial began. The whole proceeding has the emotional maturity of a group chat that should have been muted years ago.

And through all of this, Anthropic sits in the middle, quietly absorbing GPUs like a very polite black hole. The company — valued at $900 billion in its last funding round — is positioning for a year-end listing, and nothing says “we’re IPO-ready” quite like securing the largest compute partnership in the history of AI from the man who called your product evil eight months ago.

👑 The Crown Verdict

This is the AI industry in 2026, distilled to its purest essence: your enemies are your landlords, your landlords are suing your competitors, and your competitors are funded by the same cloud providers who also fund you. It’s a ouroboros of venture capital, legal filings, and Nvidia H100s, and it is magnificent.

The Anthropic-SpaceX deal isn’t just a compute partnership. It’s proof that in the race to artificial general intelligence, ideology is the first thing that gets optimized away. Musk will call your AI evil on Monday and lease you a supercomputer on Tuesday, because the real enemy isn’t the AI that might destroy humanity — it’s the other guy’s AI that might destroy humanity.

And Anthropic? They’ll take the GPUs from literally anyone, thank you very much. Safety-focused AI development requires enormous amounts of compute, and if that compute comes wrapped in a bow of interpersonal billionaire warfare, well, that’s just the cost of doing responsible business in an irresponsible industry.

The agentic era is here. The compute wars are real. And the most entertaining alliance in technology history was forged not out of shared values, but out of a shared desire to watch Sam Altman’s stock price do something unfortunate.

Inspired by LIVE: Anthropic and Elon just teamed up to take down OpenAI by Alex Finn.

Your compute allocation is showing. Lease wisely.