SpaceX Enters $40 Billion Apollo Deal for the Largest GPU Financing Ever

Starbase, Texas — SpaceX has made a binding commitment to lease approximately $40 billion of GPUs and related equipment from a vehicle run by Apollo Global Management, under a master finance lease that will supply the space and artificial intelligence company with NVIDIA's next-generation Vera Rubin chips for its data centers on the ground.

SpaceX described the commitment as the largest committed GPU financing and called the lease an important step in its plan to build a “Super Intelligence” future. The company said the deal also underscores the strength of its infrastructure capabilities.

The mechanics are set out in a Form 8-K filed with the U.S. Securities and Exchange Commission. Under the planned arrangement, the Apollo-managed lessor will buy graphics processing unit pods and related equipment and lease them to SpaceX for use at data centers the company chooses.

Deliveries will come in tranches, and each tranche gets its own lease schedule running roughly five to five and one-quarter years from its start date. Rent is due quarterly in advance. It will generally combine one-twentieth of the tranche's financed cost with a fixed annual rate on the outstanding balance, plus certain ongoing costs and expenses.

That fixed rate has not been set yet. It will be determined at pricing, which takes place before SpaceX formally enters the master lease. SpaceX puts the combined principal it will repay across all the schedules at approximately $40 billion, the amount behind the headline figure.

Apollo and Goldman Sachs lead the group of financing parties backing the deal, with Goldman Sachs acting as sole structuring agent. Funds and accounts managed by BlackRock and Goldman Sachs Asset Management also made anchor commitments, as did capital accounts advised by KKR.

The transaction is committed but not yet closed. SpaceX's entry into the lease still depends on certain specified closing conditions, and the company expects it to close before the end of 2026. The lease schedules are expected to start once the GPUs and related equipment are delivered, in the first half of 2027.

Bret Johnsen, chief financial officer of SpaceX, said: “SpaceX aims to deploy the largest compute infrastructure in the world. As demand for compute continues to accelerate, we’re seeing less than one year payback periods on our SI infrastructure investments. This lease diversifies our funding sources and addresses a notable portion of the funding required to deliver on our growth plans as we aim to get additional compute online at record speeds to power Grok for enterprises and consumers, as well as serve existing and future cloud services customers.”

The announcement follows SpaceX's third-quarter results, released on October 20. Revenue rose 161% year over year to $14.6 billion, and the company reported net income of $4.4 billion. Its Super Intelligence (SI) segment brought in $7.6 billion, which SpaceX attributed mainly to new cloud services agreements, the acquisition of Cursor and wider Grok adoption among enterprise customers and consumers.

SpaceX completed its Cursor deal on August 14, at an implied equity value of $60 billion. The company said the purchase advances its strategy of vertically integrating compute infrastructure, SI models and applications.

The compute build-out is where most of SpaceX's money is going. SI capital expenditure was about $16.44 billion in the third quarter, after $15.83 billion in the second, and reached nearly $40 billion over the first nine months of 2026. Nameplate compute capacity stood at 1.7 GW at the end of September, up from 1.4 GW at the end of June and 0.6 GW a year earlier, as SpaceX continued building COLOSSUS II.

SpaceX also raised fresh equity this year. Its initial public offering closed on June 15, when it sold approximately 638.9 million Class A shares at $135 each and raised approximately $85.7 billion in gross proceeds.

The hardware SpaceX is lining up is NVIDIA's newest. In May, NVIDIA said Vera Rubin was ramping into full production, with production shipments set to begin this fall. The chipmaker says the platform delivers 10 times the agent throughput at scale of its previous-generation Grace Blackwell systems.

Leasing the chips could let SpaceX spread a large share of the cost of its next GPU generation over about five years instead of paying for it upfront, which may leave more of its cash for other uses. It also adds fixed quarterly payments, and SpaceX lists its future debt levels and its ability to obtain financing on acceptable terms among the risks to its outlook.