SpaceX Secures $40 Billion GPU Lease Financing from Apollo for AI Infrastructure Buildout

SpaceX has finalized a binding commitment for approximately $40 billion in lease financing for AI chips and data center equipment, disclosed in a regulatory filing this week. Apollo Global Management will purchase the hardware — including Nvidia's next-generation Vera Rubin GPUs — and lease it to SpaceX over five to five-and-a-quarter years, with deliveries and lease commencements starting in the first half of 2027.

Why This Matters for GPU Infrastructure

The structure sidesteps traditional debt markets at a time when lenders have grown wary of pure-play AI compute ventures. SpaceX CFO Bret Johnsen noted the lease diversifies funding sources and covers a significant portion of the capital required for the company's AI growth plans, citing payback periods of less than one year on infrastructure investments. For the broader GPU cloud ecosystem, the deal underscores a widening divide: operators with diversified, investment-grade revenue streams (Starlink, launch services) can access cheap, long-dated capital, while the 200-plus neoclouds reliant solely on GPU rental revenue face tightening financing conditions and likely consolidation.

SpaceX's AI segment — spanning the Grok chatbot, Cursor coding assistant, and cloud compute — generated $7.6 billion of the company's $14.6 billion in Q3 revenue, surpassing Starlink and launch combined. Capital expenditures surged to nearly $20 billion year-to-date, with another $35–40 billion expected in Q4 alone. The Apollo lease provides a predictable, asset-backed funding vehicle that matches equipment useful life to cash-flow generation, a model that may become a template for other large-scale AI infrastructure buyers.