SpaceX Lands $40B for GPUs From Apollo and Predicts Extinction for Neoclouds

SpaceX has locked in a binding commitment for roughly $40B in financing, which will go toward AI chips and related data center equipment. The company disclosed the deal in a regulatory filing on Monday and called it the "largest committed GPU financing." The counterparty is Apollo Global Management, the New York private equity giant. Apollo is not lending to SpaceX directly. It buys the hardware itself and leases it out.

Leasing instead of debt: how the Apollo structure works

The mechanics are simple. A vehicle managed by Apollo buys the chips and server equipment, then leases them to SpaceX. Each lease starts when its tranche is delivered, so payments are tied to hardware actually arriving, not to the signing date. According to Apollo, the money will help get next-generation Nvidia Vera Rubin chips to SpaceX for its "terrestrial compute buildout."

Key terms of the deal:

• size: about $40B;
• format: Apollo's vehicle buys the equipment and then leases it to SpaceX;
• term of each lease: roughly five to five and a quarter years from tranche delivery;
• closing: expected by the end of the year, subject to conditions;
• lease start: first half of 2027, as GPUs and equipment arrive;
• hardware: Nvidia Vera Rubin chips and related server systems;
• stated payback on AI infrastructure: under a year, according to CFO Bret Johnsen.

Johnsen stressed that leasing "diversifies funding sources" and covers a significant portion of the funds needed for the company's growth plans. SpaceX gains another source of capital beyond traditional debt markets.

AI already earns more than Starlink and launches combined

The third-quarter numbers explain the deal. SpaceX's AI segment includes the Grok chatbot, the Cursor coding tool and a cloud compute business. It beat investor expectations, bringing in $7.6B of the $14.6B in total revenue. That is more than Starlink and rocket launches delivered combined.

Spending is growing even faster. Capital expenditures reached nearly $20B, against $6.9B a year earlier. On an investor call on October 20, Johnsen said SpaceX expects to spend $35–40B in the fourth quarter alone. By those figures, the $40B lease is comparable to a single quarter of spending. That shows the scale of purchases the company is planning.

Musk's victory lap over neoclouds

The deal came at a difficult time for AI infrastructure debt markets. Investors are nervous about runaway capital expenditures and modest profits. SpaceX had no trouble raising money, and Elon Musk took advantage of the moment. On Sunday he wrote that debt markets have effectively closed to companies that lack large sources of revenue outside AI data centers. He built his argument on SpaceX's "investment-grade" rating and the steady cash flow from Starlink and launches.

By his estimate, this threatens more than 200 neocloud companies built solely to rent out GPUs for AI workloads. "This means 200+ neoclouds are headed for a guppy-level extinction and consolidation," Musk wrote on X. "Some will survive, but not many."

The forecast remains the opinion of an interested party: SpaceX sells cloud compute itself and competes with these players. But the logic is clear. When lenders have to choose, they prefer a borrower with diversified revenue over an operator whose only asset is a fleet of rented-out GPUs.

For anyone building or renting compute capacity, a few practical takeaways follow. Equipment leasing through financial investors is becoming a workable alternative to bonds and bank loans, especially for companies with strong operating cash flow. Neocloud operators should stress-test their models early: a hardware lease of five-plus years has to be matched against customer contracts and the pace at which chip generations turn over. Finally, SpaceX's stated payback of under a year rests on its own internal demand from Grok and Cursor. Independent providers have no such in-house customer to draw on.