The Compute Utility Buys a Storefront: SpaceX, Cursor, and the Real Fight for Vibe Coding

FROM THE EDITORS

Ten days ago, we floated a heretical theory: that xAI never really cared about Grok, that its destiny was to become the “ExxonMobil of Compute” — a backend utility selling raw inference while letting someone else handle the messy consumer frontend. We even sketched the endgame: open X one morning and find a button that says “Powered by Claude.”

We got the direction right and the details spectacularly wrong.

This morning, SpaceX — fresh off the largest IPO in history — announced it is buying Anysphere, the maker of Cursor, for $60 billion in all-stock. Musk didn’t license a frontend. He bought one. And not just any frontend: the single most lucrative application built on top of large language models today.

The utility, it turns out, didn’t want to rent a storefront. It wanted to own one.

We Said “Powered by Claude.” He Bought a Claude-Killer.

Here was the logic of our June 6 piece: consumer chatbots are fickle, politically radioactive, and expensive to defend. Compute is the real moat. So xAI would quietly retreat from the frontend, plug a best-in-class model into X, and get rich selling Colossus capacity to everyone else — including, awkwardly, Anthropic.

Half of that thesis just got vindicated and half just got inverted.

Vindicated: this was never about winning the chatbot war with Grok. If it were, $60 billion would have gone into making Grok beat ChatGPT at chat. It didn’t. The capital went to a coding tool. Whatever Musk says about Grok on stage, capital allocation is the real tell — and the capital says Grok the chatbot is not the crown jewel.

Inverted: instead of partnering with the frontend leaders, Musk just declared war on them. Cursor competes directly with Anthropic’s Claude Code and OpenAI’s Codex. Our imagined “Powered by Claude” button has become something far stranger — Claude’s single biggest distribution channel is now owned by Claude’s fiercest rival.

Follow the Money, Not the Chatbot

Why coding? Because coding is the one place where AI has already turned into serious revenue. Cursor reportedly crossed roughly $2.6 billion in annualized business revenue, with enterprise sales climbing fast — making it, by some measures, the fastest-growing software company ever recorded.

That reframes the whole deal. A pure utility sells power to whoever shows up. But Musk just bought the single largest, fastest-growing consumer of his own grid. Every Cursor keystroke is inference demand — and now that demand runs on Colossus, the gigawatt supercluster in Memphis we wrote about last week.

This is forward integration, plain and simple. ExxonMobil doesn’t only pump crude; it owns the refineries and the branded stations where the margin actually lives. Musk just bought the station. The power plant and the storefront are now the same company — the grid and its most profitable customer, vertically fused.

Put it as a slogan: OpenAI owns the brand. Anthropic owns the brain. Musk owns the power — and just bought the storefront.

And here’s the elegant part for a “free speech absolutist” who keeps getting dragged into court: a coding tool generates almost none of the content-moderation, defamation, and election-integrity headaches that make a consumer chatbot a regulatory piñata. Musk found a frontend that is wildly lucrative and politically quiet. That is the most on-thesis move he could have made.

So, Who Is the King of Vibe Coding?

This is the question everyone is actually asking, so let’s answer it. The crown is a three-way fight, and each contender wins a different category:

OpenAI’s Codex has reach. It rides the largest consumer AI brand on earth and the deepest enterprise sales motion. Distribution is its weapon.
Anthropic’s Claude Code has the brain. Anthropic’s models remain the developer favorite for hard, agentic, multi-file work, and Claude Code’s terminal-native approach defined the “give the agent a task and walk away” paradigm. Quality is its weapon.
Cursor (and its Composer agent) has the users. It is the incumbent editor for an enormous base of working engineers — and as of this morning, it also has effectively unlimited compute and a two-trillion-dollar balance sheet behind it. Distribution plus capital plus compute is a frightening combination.

Add the quiet incumbent — Microsoft’s GitHub Copilot — and you have the real board. But on paper, the team that just bolted a rocket company’s war chest and a private power grid onto the most-used coding tool in the world has the strongest hand.

On paper.

The $60 Billion Catch

Here is the objection we’d raise if we were on the other side of this trade — and it’s a big one.

Cursor’s moat was neutrality. A huge part of why developers love it is that it never forced a model on them. You could run Claude inside Cursor. You could run GPT. You could pick the best brain for the job, and for hard problems, that brain was very often Anthropic’s. Cursor won, in no small part, by being the best place to use someone else’s model.

Now the best place to use Claude is owned by Claude’s competitor.

If the new owner respects that neutrality, Cursor keeps its base and $60 billion looks cheap. If it forces Grok models to the front of the queue to justify the price — and to feed an ego or two — it risks doing to Cursor exactly what acquisitions so often do: strangling the thing that made the target worth buying. The fastest-growing software company in history could become a cautionary tale about the difference between owning a product and owning its users.

That tension — neutrality versus vertical control — is the whole ballgame. Whoever is crowned king of vibe coding will be decided less by model benchmarks than by whether Musk can resist the urge to put his thumb on the scale.

The Utility Was Right

So, was our “Compute Utility” thesis wrong? No — it just leveled up.

We argued the value was migrating from the chatbot to the infrastructure. The Cursor deal confirms it and extends it: the value lives at the two ends of the stack — the power grid at the bottom and the highest-margin application at the top — and Musk now owns both. The action moved from chat to code, and from licensing the frontend to owning it.

Grok may keep its name on X for a while yet. But the question we asked ten days ago — is xAI quietly walking away from the consumer chatbot? — got a $60 billion answer this morning. Musk just told you exactly which frontend he thinks is worth fighting for.

It isn’t the one with the chat box.

Sources & further reading