SpaceX announced on the 26th that it has entered into a binding agreement with an entity managed by Apollo Global Management for an approximately $40 billion master finance lease covering GPUs and associated equipment. Under the deal, the company will deploy Nvidia’s next-generation "Vera Rubin" chips.
According to a Form 8-K filed with the U.S. Securities and Exchange Commission (SEC), the commitment letter is dated the 24th of this month, with definitive agreements expected to be finalized by the end of the year. The GPUs are set to be delivered in multiple tranches, with lease terms running for approximately five years to five years and three months from their respective inception dates. Lease payments will be made quarterly in advance, consisting of one-twentieth of the funded amount plus fixed interest on the outstanding balance, with the final interest rate to be determined prior to executing the master lease.
The financing is co-led by Apollo and Goldman Sachs, with Goldman Sachs serving as the sole structuring agent. Funds managed by BlackRock, Goldman Sachs Asset Management, and capital advised by KKR have also joined as anchor investors. SpaceX stated that the GPUs will be deployed across terrestrial data center networks rather than in space, dedicated to powering its "Grok" AI and supporting cloud service customers.
SpaceX Chief Financial Officer Bret Johnsen noted in the release that "the payback period for investments in SI (superintelligence) infrastructure is under one year," adding that the lease facility also serves to diversify the company's funding sources.
According to SpaceX’s Form 10-Q filed on the 22nd of this month, the company held a combined $38.844 billion in long-term debt—comprising long-term borrowings and senior notes—and capital lease obligations (leases that effectively function as installment purchases) as of September 30. At approximately $40 billion, the principal of this new lease facility exceeds the sum of all its existing liabilities in a single transaction. If fully drawn, the principal amortization portion of the lease payments alone would come to roughly $2 billion per quarter, or about $8 billion annually.
In its third-quarter results reported on the 20th of this month, revenue from the SI division—which drives SpaceX's AI initiatives—reached $7.560 billion, accounting for more than half of total consolidated revenue of $14.575 billion. From a top-line perspective, AI has already become the company’s core business. On the bottom line, however, heavy upfront investment continues to dominate: the division’s capital expenditures totaled $16.444 billion, more than double its revenue, leaving operating income at just $431 million. That is roughly one-fifth of the $2.367 billion in operating profit generated by the Connectivity division, which includes Starlink.
Spending is poised to accelerate even further. During the earnings call, Johnsen guided fourth-quarter capex to between $35 billion and $40 billion. That would represent a 1.8- to 2-fold increase over the $19.593 billion in total capex recorded across the company in the third quarter.
SpaceX CEO Elon Musk indicated in a post on X on the 25th of this month that the financing environment for AI is tightening rapidly. "In the last few weeks, debt markets have almost completely stopped funding companies that don't have substantial revenue independent of AI data centers," he wrote. His remarks highlighted that financial institutions and investors have ceased extending credit to companies lacking reliable revenue streams outside of their AI data center operations.
Addressing cloud providers dedicated specifically to leasing GPUs (known as "neoclouds"), he added, "That means the 200+ neoclouds are going to see guppy-level death and consolidation. Some will make it through, but not many." His assessment suggests that drying liquidity will push many of these neoclouds into insolvency or forced consolidation. Dismissing speculation that SpaceX might acquire such struggling providers, Musk flatly denied interest, replying: "SpaceX will not buy neoclouds. Makes no sense."
In the same thread, Musk claimed the $40 billion financing had "closed." However, Bloomberg reported, citing people familiar with the matter, that negotiations over pricing terms—including interest rates—were still ongoing as of the evening of the 25th. SpaceX's Form 8-K likewise stated that entering into definitive agreements remains subject to specified closing conditions, meaning the transaction is not yet finalized.
The financing structure has also evolved from initial media reports. On the 6th of this month, the Financial Times reported that SpaceX was in discussions for an Apollo-led financing package of around $40 billion. At the time, the deal was envisioned as SpaceX borrowing the funds directly to purchase the chips, backed by roughly $10 billion in bank facilities and $30 billion in investment-grade debt. What was officially announced, however, is a structure under which an Apollo-managed entity purchases and holds title to the GPUs, leasing them to SpaceX.
Precedent exists for this model of a third party buying chips and leasing them to the end operator: a facility engineered by Apollo and Blackstone for Anthropic. In that transaction, a special-purpose vehicle (SPV) purchases Google’s Tensor Processing Units (TPUs) and leases them to Anthropic, with Broadcom reportedly guaranteeing the residual value of the silicon at the end of the lease term. The deal was estimated at around $36 billion when first reported in May, before reportedly closing at roughly $35 billion in June.
Image: Shutterstock
SpaceX Earnings: AI Revenue Overtakes Starlink as Company Posts First Profit Since IPO
The Invisible Credit Behind the AI Market—Who Funded the Purchase of GPUs?
