Musk Once Slammed Anthropic As A Threat to Civilisation. Now He Powers It

In politics, they say there are no permanent enemies, only permanent interests. In the Silicon Valley of 2026, it turns out there are no permanent foes, just a permanent, desperate hunger for more GPUs.

The tech world woke up this week to a partnership that feels like a glitch in the simulation: Elon Musk is officially the landlord for Anthropic.

But wait, what's a GPU? Think of a standard computer chip (CPU) as a brilliant professor who solves one complex problem at a time. A GPU (Graphics Processing Unit) is like a stadium full of math students all solving thousands of simple problems simultaneously. In the world of AI, you need that "stadium" speed to process the massive amounts of data required to make a chatbot sound human.

So, this deal is a massive strategic pivot. Anthropic is taking over the entire capacity of Colossus 1, the Tennessee-based supercomputer owned by SpaceXAI (the newly merged giant of xAI and SpaceX). We are talking about over 220,000 NVIDIA GPUs being handed over to the very company Musk has spent the last year trashing on X (formerly Twitter).

For a man who spent much of 2025 and early 2026 calling Anthropic's Claude "misanthropic," "evil," and a "threat to Western civilisation," the U-turn is stunning. But then it's Musk after all who has built a reputation of being this unpredictable tech genius, often dubbed the real-life Tony Stark/ Iron Man. In fact, he has been credited by director Jon Favreau for serving as an inspiration for the on-screen fictional character played by Robert Downey Jr. Musk even made a cameo in Iron Man 2 as himself coming face-to-face with Tony Stark's character.

Coming back to the current development, in the AI arms race, 300 megawatts of power is a much better olive branch than an apology. Anyway, an apology from Musk would seem really out of character.

The 'Evil Detector' And The Kill Switch

This isn't your standard corporate handshake. It's a deal with a very Musk-ian sword of Damocles hanging over it. Musk didn't exactly offer a warm welcome; instead, he sounded like a landlord warning a tenant not to burn the house down.

"I spent a lot of time last week with senior members of the Anthropic team... No one set off my evil detector," Musk posted on X.

But here's the kicker: the contract allegedly includes a "Humanity Clause." SpaceXAI reportedly reserves the right to reclaim the capacity if Claude "engages in actions that harm humanity."

Essentially, Musk has his finger on the plug. If Claude leans too far into the "woke" or "biased" territory that Musk frequently rails against, he can evict them from the data centre in a heartbeat. It's essentially a partnership built on a hostage situation, except the hostage is paying a massive amount of rent.

The Cold Calculus Of Compute

So, why the sudden change of heart? For Anthropic, the reason is simple: hunger and growth.

Despite a nearly $400 billion valuation (pegged to go up to $900 billion+ soon) and backing from Google and Amazon, Anthropic has been hitting a wall. Their models are beloved, but their users were constantly hitting rate limits. To keep pace with OpenAI, they needed "atoms moved" fast. As Anthropic's leadership noted, SpaceX is the only organisation that can deploy infrastructure at this "unprecedented scale" on a timeline that actually matters.

For Musk, the deal is pure pragmatism. SpaceXAI has already moved its own flagship training for Grok to the even more powerful Colossus 2. Leaving Colossus 1 idle would be a sin against his productivity metrics. By leasing it to a rival, he turns a massive overhead into a massive revenue stream, all while keeping his competitor on a very short, very high-voltage leash