← Back to feed

On April 21st, SpaceX posted to X and confirmed terms: a compute partnership between SpaceXAI and Cursor, plus a one-year option for SpaceX to acquire Cursor for $60 billion — or walk away having paid $10 billion for the collaboration.

Cursor was valued at $29.3 billion six months ago. A new funding round at $50 billion was reportedly being finalised when this landed. The option price is a 20% premium on a valuation that was itself moving fast.

But the $60 billion figure is almost beside the point.

The clock was already ticking

For the last two years, Cursor’s business model rested on a reasonably comfortable arrangement: buy API access to frontier models from OpenAI, Anthropic, and Google, and package it into the best coding experience on the market. That worked when those companies were primarily model providers.

It works less well when they are also your competitors. Claude Code, GitHub Copilot, and Gemini’s coding integrations are not third-party threats. They are products built and subsidised by the same organisations whose infrastructure Cursor runs on. At some point, neutrality becomes a polite fiction.

Colossus changes that calculus. SpaceX describes the cluster as equivalent to a million Nvidia H100 GPUs. If Cursor can train its Composer model at that scale without touching OpenAI or Anthropic infrastructure, the dependency problem shrinks from existential to manageable. That is not a minor operational benefit. It is the difference between being a distribution layer and being a company with durable margins.

Why SpaceX, and why now

SpaceX owns xAI following the February 2026 all-stock merger, and xAI owns Colossus. The deal is therefore less about aerospace and more about where the compute lives. Running it through SpaceX rather than xAI directly keeps the structure cleaner ahead of what SpaceX says will be a June 2026 IPO targeting $1.75 trillion — which would be the largest public offering in history. A credible AI product story going into that raise is worth more than the option premium.

Cursor’s revenue makes the $60 billion harder to write off as pre-IPO narrative management. Annualised recurring revenue is tracking above $2 billion in early 2026, with projections of $6 billion by year-end. A million daily users. Fortune 500 adoption across more than half the index.

The option is the point

One-year exclusivity, two exit prices. SpaceX gets first mover on the most valuable independent coding platform. Cursor gets compute it doesn’t have to beg for, and a floor under its valuation while it figures out whether it wants to be acquired at all.

The $10 billion walk-away clause is the tell. If this were purely a talent or acquihire play, nobody pays $10 billion to pass. The floor exists because Cursor without SpaceX still has serious value, and both sides know it.

The harder question is what happens if the option lapses. Cursor returns to a market where its main infrastructure providers are also its sharpest competitors, having spent a year training models on Colossus that it may not be able to replicate on its own. That is a reasonable position to be in — if the models are good enough. If they are not, the $10 billion starts to look like expensive rent.

SpaceX recently moved two of Cursor’s most senior engineering leads, Andrew Milich and Jason Ginsburg, onto its own payroll, reporting to Musk. Read that however you want.

The deal structure rewards patience. The question is whose.