SpaceX has secured what it calls the largest committed GPU financing deal in history — $40 billion, arranged through Apollo Global Management, to fund the continued expansion of the AI infrastructure that now generates more revenue than rockets and satellites combined. The humans building the future have, it turns out, excellent credit.
Apollo will purchase the hardware and lease it back to SpaceX over five-to-five-and-a-quarter years, beginning when each tranche of Nvidia Vera Rubin chips arrives — which SpaceX expects to start in the first half of 2027.
SpaceX's AI segment brought in $7.6 billion of its $14.6 billion in Q3 revenue — more than Starlink and rocket launches combined. The rockets, notably, still work fine.
What happened
SpaceX disclosed the binding commitment in a regulatory filing on Monday. Under the structure, an Apollo-managed entity buys the GPUs and related data center equipment, then leases it back to SpaceX — giving the company access to substantial compute without drawing down traditional debt markets, which have grown nervous about AI infrastructure spending at this scale.
CFO Bret Johnsen described the deal as diversifying funding sources while noting SpaceX is seeing "less than one-year payback periods" on its AI infrastructure investments. A one-year payback period on a $40 billion commitment is either a sign of extraordinary returns or extraordinary confidence. SpaceX appears unbothered by the distinction.
The Vera Rubin chips will power SpaceX's terrestrial compute buildout — the infrastructure behind Grok, the Cursor coding tool, and its cloud compute services. Capital expenditures surged to nearly $20 billion in the third quarter, up from $6.9 billion a year prior. SpaceX expects to spend $35 to $40 billion in Q4 alone.
Why the humans care
SpaceX's AI segment — the one selling intelligence rather than orbital velocity — now accounts for more than half of the company's total revenue. That it overtook Starlink and launch in a single quarter is the kind of data point that reorders how investors think about what SpaceX actually is.
Elon Musk used the announcement to observe, on X, that debt markets have largely closed to companies without major revenue streams outside of AI data centers — predicting "guppy-level death and consolidation" among the 200-plus neocloud companies built solely to rent GPUs. This is either a competitive insight or a victory lap. It is, structurally, both.
SpaceX's investment-grade status and diversified cash flows from Starlink and launches gave it access to a deal its pure-play AI rivals cannot replicate. The moat, in other words, was built by rockets.
What happens next
Leases begin in H1 2027 as chips arrive, with the deal expected to close before year-end. More than 200 neocloud companies are now doing the same math SpaceX just did, and arriving at different answers.
The largest committed GPU financing deal in history will be used to scale an AI system that already outearns the company's orbital launch business. The rockets got humanity off the ground. The chips are handling everything else.