SpaceX says it has secured $40 billion in debt financing for GPUs due in the first half of 2027, a sizable bet on AI computing capacity backed by revenue from Starlink and its launch business. Elon Musk announced the financing on X on October 25, agreeing with investor Chamath Palihapitiya’s warning that data center delays and tighter funding could leave some AI hardware without a place to run. The deal links two parts of SpaceX’s business: established revenue that can support borrowing, and a large expansion into computing that depends on infrastructure being ready when the hardware arrives.
Musk said debt markets had largely stopped lending to companies without significant revenue outside AI data centers. SpaceX, he argued, has an advantage because it has investment-grade status and established income streams. That distinction matters in a sector where companies may have access to chips but lack the cash flow or facilities needed to put them to work. He predicted that many of the more than 200 “neoclouds” would face severe distress and consolidation, though he said some would survive.
Financing details and timing
The word “closed” leaves some details unresolved: Musk hasn’t said whether final debt pricing is set, or specified the GPU models and quantities covered. Those details affect how readers should interpret the headline figure: a financing commitment is a major signal of intent, but it doesn’t by itself establish the final cost or the precise scale of the hardware purchase. He described the borrowing as primarily callable debt for purchases expected to arrive in the first and second quarters of 2027, adding that it wouldn’t be a major factor in the current quarter.
The financing follows reports earlier this month that SpaceX was seeking a $40 billion package for Nvidia hardware. That proposed package was described as $10 billion in bank loans and $30 billion in investment-grade debt, potentially covering about 360,000 Rubin GPUs across 5,000 Nvidia NVL72 racks. Musk has confirmed the financing, but not those reported hardware figures. The distinction is material: the reported structure and equipment list offer a possible explanation for the amount, but they remain unconfirmed details rather than a disclosed final order.
The commitment comes as SpaceX prepares for a large spending period. CFO Bret Johnsen said on the company’s Q3 2026 earnings call that capital expenditure in Q4 could reach $35 billion to $40 billion, citing additional capacity coming online during the quarter. That spending plan gives the GPU borrowing a broader context. SpaceX is planning for substantial investment across its business, so the new debt sits alongside a period when the company expects to spend heavily on capacity.

- The data center cooling state of play
- The custom AI ASIC state of play
- America’s AI chip rules keep changing — and the rest of the world is paying the price
- GTC 2026: Ian Buck press Q&A transcript — VP of Hyperscale and HPC speaks out on shelving CPX and shipping LPU decode this year
- Demand for data center CPUs has surged, and AI agents are responsible
Compute plans extend beyond this purchase
The GPU financing sits within a broader expansion effort. In August, xAI said it planned to increase data center capacity sevenfold by 2027, targeting 10 gigawatts. A September report said SpaceXAI planned to add 660,000 Nvidia GPUs during 2026, which could bring its total to about 1.44 million if deployments proceed. These plans suggest the financing is one piece of a larger buildout, though reported targets don’t establish that the equipment or capacity will be delivered on schedule.
Getting accelerators is only one part of building usable AI capacity. Power, cooling, networking, and data center construction all have to keep pace, a dependency that makes the financing’s delivery timeline only one measure of progress. A GPU shipment can be counted before a site is ready to run it; the business value comes later, once the surrounding systems are operating and customers or internal workloads can use the machines. And this is the part most people overlook: purchasing hardware can move faster than completing the facilities that make it productive.
SpaceX has also begun selling access to its computing infrastructure. In May, Anthropic secured access to Colossus 1, a supercomputer with more than 220,000 Nvidia GPUs and over 300 megawatts of capacity. In June, Google agreed to pay $920 million per month for access to about 110,000 Nvidia GPUs and related hardware, with payments scheduled from October 2026 through June 2029. Those arrangements show how existing capacity can produce revenue while a company expands, but they don’t answer whether the next wave of planned systems will be ready when customers need it.
What the financing could mean
My read is that SpaceX is positioning itself on both sides of the AI compute market: buying capacity for its own expansion while renting out existing systems. This is more significant than it looks because the company’s argument for borrowing rests partly on revenue beyond AI, while its growth plans increasingly depend on AI infrastructure. If you're working in this space, the distinction between financing hardware and bringing usable capacity online is the one to watch.
That said, the available details leave open basic questions about the debt’s final terms, the GPU order, and the timing of deployment. SpaceX may be better placed than a company relying solely on AI data center revenue to secure financing, but that advantage doesn’t remove the physical constraints of building and powering facilities. The test comes when deliveries meet infrastructure. For now, a large commitment—and a demanding execution schedule.