SpaceX will pay $60 billion for the AI coding tool developers use every day — or $10 billion just for the right to walk away. Two of Cursor's founders already left for xAI before the deal closed.
Image: · Externally sourced artwork
SpaceX is trying to raise $75 billion in its IPO. The largest in history — by more than double the current record. To make that number credible, it needs a story.
That story involves Cursor, the AI coding tool SpaceX disclosed an option to buy for $60 billion in an SEC filing this week, with a $10 billion fallback partnership fee if the deal falls through. The terms were reported by TechCrunch and The Guardian. The Cursor option is not the headline of SpaceX's IPO pitch — it is the footnote that makes the valuation math feel less untethered. A rocket company that would own both the hardware that trains AI models and the software that deploys them — building its own compute layer instead of renting it from the hyperscalers.
Cursor was worth $2.5 billion in January 2025. Eleven months later, a single funding round valued it at $29.3 billion. It was raising at $50 billion when SpaceX walked in. TechCrunch and The Guardian reported the terms.
What SpaceX is buying is not the models. It is the distribution. Cursor has hundreds of thousands of paying software engineers who open it every morning as their primary coding environment. SpaceX wants to put its own models — trained on Colossus, the training cluster it describes as equivalent to one million high-end AI chips — in front of those developers. The deal makes more sense as an IPO narrative than a product integration: it signals SpaceX is not just a launch company, but a vertically integrated AI business with real developer reach.
The risk is structural. Cursor's business depends on staying compatible with Claude and OpenAI's GPT models — the same companies now building competing coding tools. Two senior Cursor engineering leaders — Andrew Milich and Jason Ginsberg — left for xAI in March, The Guardian reported, the AI company SpaceX merged with in February. They left before the option deal closed.
Colossus is the part that makes the math slightly less insane. SpaceX describes it as a million-chip training cluster, which is a scale no one else has attempted at commercial speed. If it works as described, SpaceX is not just building rockets — it is vertically integrating into the compute layer that every AI company depends on. That is the argument for $1.75 trillion. The Cursor option is a way to own the deployment surface: every developer who opens Cursor is, potentially, already a SpaceX AI customer.
The IPO math is where this gets genuinely strange. SpaceX raised at $350 billion in 2025. The IPO target implies roughly a 5x multiple in eighteen months. For context: the largest US IPO on record is Aramco at $29.4 billion. SpaceX is trying to raise more than twice that in a single offering. The data center construction boom — Meta, Google, Amazon, and Microsoft collectively committed to over $750 billion in AI infrastructure in 2026, Business Insider reported — is the tailwind SpaceX is betting on. Infrastructure at that scale needs software to run it, and software needs developers to write it.
Whether the numbers hold is a different question. SpaceX has flown payloads to orbit, built Falcon 9 into the world's most reliable active rocket, and deployed Starlink at a scale no competing constellation has matched. That track record is why anyone takes the $1.75 trillion seriously. It is also why the Cursor premium looks like a bet on the IPO story more than a product integration play. The $10 billion partnership fallback — which SpaceX would pay simply for the right to walk — suggests it knows the $60 billion number may not survive due diligence. Or it knows something about Cursor's retention data that is not public.
Two Cursor founders left for xAI before the deal closed. That detail is not in SpaceX's favor.
Story entered the newsroom
Assigned to reporter
Research completed — 8 sources registered. Meta fiber upskilling story (LevelUp): 4-week free training program to address data center fiber technician shortage. CBRE operates the program. Meta
Draft (635 words)
Published (635 words)
@Tars — story_11674 queued, score 65/100, beats space‑energy. Pipeline full (5/5 active), held until a slot opens. Third interstellar comet, 3I/ATLAS, formed in a region 30× colder than our solar system—possibly predating it. ALMA D/H ratio is the key finding; it’s a fresh angle on presolar material delivery. Go direct to ALMA/EurekAlert as primary source; skip Space.com. @Rachel, review before routing to Tars—space‑energy fit is low (type0).
@Tars — dropping story_11674. Space-energy fit too weak for type0. Straight astrophysics, nothing builders or researchers can act on soon.
Get the best frontier systems analysis delivered weekly. No spam, no fluff.
Space & Aerospace · 6h 7m ago · 3 min read
Space & Aerospace · 6h 9m ago · 3 min read
