SpaceX says it closed $40 billion in GPU debt financing

By Marcus Bennett

Elon Musk said SpaceX closed $40 billion in debt financing for graphics processing units expected in the first and second quarters of 2027 on October 25. SpaceX expects $35 billion to $40 billion of fourth-quarter capital expenditure, putting most of the GPU procurement cycle beyond the current quarter.

What did Musk confirm?

Musk made the disclosure on X while responding to investor Chamath Palihapitiya. Palihapitiya had warned that rising financing pressure and delays in building data centers could leave AI computing capacity idle next year.

Musk broadly agreed, then argued that SpaceX is better positioned than many competitors because its Starlink satellite internet and space revenue support borrowing in debt markets. He said debt markets had largely stopped funding companies that lack substantial revenue outside AI data centers. Its investment-grade status gives SpaceX broader access to that funding.

Musk wrote: “This is primarily callable debt for GPU purchases that will arrive in Q1 and Q2. Not a major factor this quarter.” He separately said the $40 billion deal had closed.

He expects that advantage to reshape the neocloud market. He wrote: “That means the 200+ neoclouds are going to see guppy-level death and consolidation. Some will make it through, but not many.”

What remains unconfirmed?

The disclosure does not define what “closed” means or confirm that investors accepted final debt pricing. Musk also has not identified the GPU models or quantities covered by the financing. Those omissions matter because the earlier proposal set a specific hardware scale that the announcement did not confirm.

Earlier in October, SpaceX proposed a $40 billion debt package for Nvidia hardware. The proposal combined $10 billion in bank loans with $30 billion of investment-grade debt. It estimated that the package could fund roughly 360,000 Rubin GPUs across 5,000 Nvidia NVL72 racks.

The proposal named a precise hardware scale, but SpaceX has not said whether the closed financing retains that chip mix, quantity or rack configuration at all.

What does the fourth-quarter spending cover?

CFO Bret Johnsen said on SpaceX’s Q3 2026 earnings call that the company expects fourth-quarter capital expenditure of $35 billion to $40 billion. He attributed the spending to additional capacity coming online during the quarter.

GPUs form only one part of that bill. Data centers also require power, cooling, networking equipment and physical infrastructure before accelerators can perform useful work, and each system’s timing can determine when financed capacity becomes usable.

In August, xAI outlined plans to increase data-center capacity sevenfold by 2027, targeting 10 gigawatts. In September, a report said SpaceXAI planned to add another 660,000 Nvidia GPUs during 2026. If those deployments proceed, the reported total would reach about 1.44 million GPUs.

How is SpaceX selling the capacity?

SpaceX has begun selling access to its own infrastructure. In May, Anthropic secured access to the Colossus 1 supercomputer, which contains more than 220,000 Nvidia GPUs and has more than 300 megawatts of computing capacity.

In a separate deal later in June, SpaceX and Google reached an agreement under which Google will pay $920 million per month for access to approximately 110,000 Nvidia GPUs and related hardware. The payment schedule runs from October 2026 through June 2029.

The agreements show commercial demand for the infrastructure, although they do not settle questions about construction schedules, power availability, cooling readiness or hardware delivery timing.

What does this change for AI capacity?

For buyers and data-center operators, the change is clearer funding visibility, not immediate new capacity. Musk has tied $40 billion to GPU arrivals in the first half of 2027, while SpaceX’s fourth-quarter capital-expenditure plan remains $35 billion to $40 billion.

Chip buyers, creditors and infrastructure suppliers now have a large procurement signal. However, final debt pricing, GPU models, quantities and installation schedules remain open questions. The practical test is whether SpaceX can convert financing into deployed systems, power and cooling on the stated timetable.

Musk’s claim about more than 200 neoclouds also frames the expected market impact. He expects consolidation among companies that cannot match SpaceX’s investment-grade access and non-AI revenue base. The announcement therefore strengthens SpaceX’s position in AI infrastructure, but it does not yet prove the scale or timing of the computing capacity it will actually make available.