SpaceX has received approval to acquire the AI startup Cursor for $60 billion

(source) SpaceX has received approval to acquire the AI startup Cursor for $60 billion (ua.news)

The Cursor-SpaceX deal looks wild: a coding startup gets Colossus-scale compute, while SpaceX gets the right to buy Cursor for $60B or pay $10B to walk. Effectively a one-year call option. The deal makes a lot of sense for one reason - and very little for another.

What Makes Sense - The structure reveals where power is concentrated.

For the last two years, AI coding looked like an app-layer land grab: build the best interface, win developers, monetize seats. Reasonable, but incomplete.

Cursor built a great product and put up one of the fastest ramps in recent history. But the market shifted. Now OpenAI has Codex, Anthropic has Claude Code, and Google has its own stack.

The labs that own the models are increasingly owning the experience. That makes being the best shell around someone else’s intelligence precarious. Cursor’s problem is simple: its suppliers have become its competitors.

Anecdotally, the market is already full of stories of teams moving off Cursor and directly onto Claude Code or Codex. The center of gravity is shifting from the wrapper → first-party system. If value accrues to the model, neutrality isn’t a moat. You need privileged access - or your own capability. Cursor tried to evolve into a full-stack intelligence company but was bottlenecked by compute.

From SpaceX’s perspective, xAI has compute but no app-layer traction. The move is obvious: turn infra into leverage. Find an app that already matters, give it the GPUs it desperately needs, and buy an option on the upside. That is a pretty good trade.

Cursor gets computed. xAI gets a real use case. Spare GPUs meet existential need. As marriages of convenience go, this one is unusually clean.

What Makes Less Sense

Cursor’s value has been model-agnostic flexibility. xAI ownership threatens that neutrality.

Anthropic has already shown that model access is not some apolitical utility. When OpenAI was rumored to be acquiring Windsurf, Anthropic cut its direct access to Claude models. That precedent hangs over this deal.

Why would Anthropic or OpenAI let their best models power a rival-owned product - while feeding it data and insight? Even with strict separation, incentives are misaligned.

So the Cursor-SpaceX deal only makes sense if one of three things is true:

1. Cursor thinks Anthropic will stay on-platform despite the ownership change, because the commercial upside is too large

2. Cursor thinks losing access to Claude is survivable

3. Cursor has decided that dependence on Claude was already an existential risk, and the only rational response was to get closer to owning more of the stack before the supplier relationship turned fully adversarial.

This is not just a startup cashing in at a huge price. It is one of the breakout winners of the AI app wave, confronting a harsher reality: in AI, the most important product in your stack can also become your most dangerous landlord.