Elon Called Anthropic 'Missanthropic' in March — Then Signed a Compute Deal With Them in April
Elon Musk publicly called Anthropic 'the most hypocritical company' in March 2026. Weeks later, SpaceX signed a major compute deal with them. Here's why.
Elon Called Anthropic ‘Missanthropic’ in March. He Signed a Deal With Them in April.
In March 2026, Elon Musk called Anthropic “the most hypocritical company” and said “winning was never in the set of possible outcomes for Anthropic.” Weeks later, SpaceX signed a deal giving Anthropic the entire capacity of the Colossus 1 data center.
That’s not a rumor. That’s the actual sequence of events.
If you’ve been following AI infrastructure news, you’ve probably seen the headlines but not the full picture. This post is an attempt to lay out the actual logic — why Elon said what he said, why he did what he did, and why those two things are less contradictory than they appear.
What Elon Actually Said (and When)
The timeline matters here.
As recently as March 2026, Musk was publicly dismissive of Anthropic. He called it “missanthropic” — a play on the company name — and framed it as a company that couldn’t win. “Winning was never in the set of possible outcomes for Anthropic” is a direct quote. He also called it “the most hypocritical company,” which is a reference to Anthropic’s safety-first positioning while operating as a well-funded commercial AI lab.
This wasn’t an old tweet from 2023. This was March. The deal came in late April or May.
Everyone else built a construction worker.
We built the contractor.
One file at a time.
UI, API, database, deploy.
So the question isn’t whether Elon changed his mind. It’s whether he had a reason to act against his stated opinion — and whether that reason was strong enough to override the public posturing.
It was.
The Compute Math That Makes This Make Sense
Here’s the number that explains almost everything: 11%.
Elon has said publicly that XAI is only using approximately 11% of its available compute. Colossus 1 — the data center SpaceX built — is sitting at roughly 89% idle capacity. That’s not a rounding error. That’s a massive, expensive infrastructure asset generating almost no return.
Meanwhile, Anthropic has the opposite problem. The company has been constrained by compute shortages for months. Users hitting Claude’s usage limits has been one of the most consistent complaints from developers and power users. Anthropic had models people wanted to use and not enough infrastructure to serve them.
So you have one party with compute and no demand, and another party with demand and no compute. The deal writes itself.
The specific terms: Anthropic takes over the entire capacity of Colossus 1. Not a slice of it. All of it. And separately, Anthropic committed $200 billion to Google Cloud over five years — a number that was reported by The Information and caused Google’s stock to spike 1.5% overnight after the backlog announcement had already pushed it up 10%.
The direct result for users: Claude Code hourly limits doubled. (Weekly limits hadn’t changed as of the reporting, but the hourly increase was immediate.)
Why ‘Enemy of My Enemy’ Is Only Part of the Story
The most common explanation you’ll see for this deal is that Elon is helping Anthropic because Anthropic competes with OpenAI, and Elon is currently in an active lawsuit with Sam Altman.
That explanation isn’t wrong. It’s just incomplete.
Yes, Elon and Sam Altman are in litigation. Yes, Elon helping Anthropic pull ahead of OpenAI would be a satisfying outcome from his perspective. The “enemy of my enemy” framing is real and probably part of the calculus.
But it doesn’t explain why Elon would hand over the entire Colossus 1 facility — not just sell some spare capacity, but give Anthropic the whole thing. That’s a much bigger commitment than spite requires.
The fuller explanation involves Terrafab.
Terrafab: The Project That Needed a Customer
In March 2026, Elon announced Terrafab — a chip manufacturing facility in Grimes County, Texas. Initial estimates put the cost at $20–25 billion. People mostly shrugged. Even with Intel added as a partner in April, analysts were skeptical. Nvidia analyst Tae Kim was blunt: “I’m not that optimistic. It’s so hard to build that. It’s almost like cooking where it takes a lot of trial and error accumulated over decades.”
The skepticism was reasonable. Building a chip fab from scratch is genuinely hard. TSMC took decades to develop its manufacturing expertise. The idea that Elon could stand up a competitive fab quickly enough to matter seemed like a stretch.
Plans first. Then code.
Remy writes the spec, manages the build, and ships the app.
Then the Anthropic deal happened, and a legal filing in Grimes County revealed the updated cost estimate: $55 billion to $119 billion. That’s not a rounding error from the original $20–25 billion estimate. That’s a completely different project.
If completed, Terrafab would be the largest chip fab on the planet.
Now the deal makes a different kind of sense. Anthropic isn’t just a customer for Colossus 1. Anthropic is the demand signal that makes Terrafab credible. Before the deal, Terrafab’s potential customer base was Tesla and Optimus — real demand, but not the kind that justifies the world’s largest semiconductor facility. After the deal, Elon has a basically unlimited source of AI compute demand from one of the two or three most credible AI labs in the world.
The Anthropic partnership doesn’t just monetize idle capacity. It provides the demand justification for an infrastructure bet that’s now in the $55–119 billion range.
What This Tells You About Elon’s Actual Strategy
XAI is ceasing to exist as a separate company. It’s being fully folded into SpaceX.
That’s worth sitting with for a moment. Elon started XAI to compete directly with OpenAI on model development. Grok exists. Grok 4.3 was released recently and made real benchmark improvements — but it’s still not in the same tier as Claude, GPT-5, or Gemini on most evaluations. The model competition hasn’t gone the way Elon probably hoped.
What has gone well is infrastructure. Colossus 1 was stood up in record time. The ability to build and operate large-scale compute facilities quickly is something Elon has demonstrated — it’s the same skill set that made Tesla’s Gigafactories work and that made Starship development move faster than anyone expected.
The AI Daily Brief framed it well: “Peak Elon was scaling up Tesla production when he famously slept on the factory floor in 2018. And maybe the closest example we’ve had recently was him standing up the first Colossus data center in record time at the end of 2024.”
Folding XAI into SpaceX and pivoting toward infrastructure is Elon playing to his actual strengths rather than continuing to compete in model development against labs that have a significant head start.
What the Google Deal Tells You About Anthropic’s Position
The $200 billion Google Cloud commitment is worth examining separately from the SpaceX deal, because it reveals something about Anthropic’s strategic situation.
Anthropic was, by most accounts, more conservative than OpenAI in its infrastructure investments. That conservatism made sense from a capital efficiency standpoint — but it meant that when demand for Claude accelerated, Anthropic didn’t have the compute to serve it. The Claude usage limits that frustrated developers were a direct consequence of that underinvestment.
The $200 billion Google commitment is Anthropic overcorrecting. It’s a massive, multi-year commitment that signals Anthropic is done being cautious about compute. Combined with the SpaceX deal, Anthropic is now sourcing capacity from two major infrastructure providers simultaneously.
Hire a contractor. Not another power tool.
Cursor, Bolt, Lovable, v0 are tools. You still run the project.
With Remy, the project runs itself.
Google’s market reaction is telling. The stock was already up 10% after the backlog announcement during earnings. It spiked another 1.5% when the $200 billion number was reported, and it held those gains through the end of the week. Analysts aren’t treating this as circular funding — they’re treating it as real demand from a company with real revenue growth.
For developers building on Claude, this matters practically. More compute means higher limits, more consistent availability, and eventually lower prices. The doubling of Claude Code hourly limits is the first visible result. It probably won’t be the last.
The Part That’s Still Confusing
One thing that doesn’t fully resolve: Elon’s public statements about Anthropic were recent and specific. “The most hypocritical company” isn’t a throwaway comment. It suggests genuine contempt, or at least a genuine desire to project contempt.
The most charitable reading is that Elon’s public statements and his business decisions operate on different tracks. He says things for effect — to shape perception, to needle competitors, to signal tribal affiliation — and then makes decisions based on different criteria. The lawsuit with Altman is real, but so is the business logic of monetizing 89% idle compute.
The less charitable reading is that Elon knew the deal was coming when he made those statements, and the public contempt was positioning — making it look like the deal was a reluctant concession rather than a strategic choice.
Either way, the outcome is the same. Anthropic gets compute. SpaceX gets revenue. Terrafab gets a customer. And Elon gets to be in the infrastructure business rather than the model business, which is probably where his actual competitive advantage lies.
What This Means If You’re Building on Claude
If you’re an AI builder, the practical implications are fairly direct.
Claude Code hourly limits doubled as a direct result of the Colossus 1 deal. If you’ve been hitting limits mid-session — which has been a real workflow problem for anyone doing extended Claude Code sessions — that constraint just got meaningfully less painful.
The $200 billion Google commitment suggests Anthropic is planning for sustained capacity growth, not just a short-term fix. That’s relevant if you’re making architectural decisions about which models to build on top of.
For teams building multi-model workflows — where you might route different tasks to different models based on cost, capability, or availability — platforms like MindStudio handle this orchestration across 200+ models, which means infrastructure shifts like this one get absorbed at the platform level rather than requiring you to rewire your integrations every time a capacity deal changes the calculus.
And if you’re thinking about what the Terrafab announcement means for the longer arc: if Elon can actually build the world’s largest chip fab with Anthropic as the anchor customer, the compute constraints that have defined the last two years of AI development start to look temporary rather than structural.
The Broader Pattern
The AI infrastructure story has a consistent shape right now: demand is growing faster than supply, supply is catching up through massive capital commitments, and the companies that can credibly claim demand are the ones that can justify the infrastructure bets.
- ✕a coding agent
- ✕no-code
- ✕vibe coding
- ✕a faster Cursor
The one that tells the coding agents what to build.
Anthropic’s $200 billion Google commitment caused a 10% stock spike not because the number is circular — Google invested in Anthropic, Anthropic pays Google — but because analysts believe Anthropic has the revenue trajectory to actually pay it. That’s a different conversation than the one happening six months ago.
The Terrafab cost revision from $20–25 billion to $55–119 billion is the same pattern from the other direction. The project got more credible as the demand signal got clearer. Anthropic is now that demand signal.
For developers building applications on top of these models, the infrastructure layer is becoming more stable, not less. The compute wars are real, but the outcome — more capacity, higher limits, eventually lower costs — is good for builders. Tools like Remy are designed around this assumption: that the underlying model infrastructure is reliable enough to compile a full-stack application from a spec, with the generated TypeScript, database, and deployment all depending on consistent model availability.
Elon called Anthropic “missanthropic” in March. He signed a compute deal with them in April. The contradiction is real, but the logic underneath it is straightforward: 89% idle compute, a lawsuit with OpenAI, a chip fab that needed a customer, and a company that plays to its strengths when it’s paying attention.
Infrastructure, not models, is where Elon’s story is heading. The Anthropic deal is the clearest signal yet.
