SpaceX is reportedly looking to raise $40 billion to finance a large purchase of Nvidia artificial intelligence chips, in what could become one of the clearest signs yet that Elon Musk’s space company is becoming a major AI infrastructure operator.
According to Reuters, citing people familiar with the matter, SpaceX is discussing a financing package that would combine roughly $10 billion in bank loans with $30 billion of investment-grade debt. The Financial Times first reported the proposed financing. Apollo Global Management is expected to lead the transaction, while bond manager PIMCO is among the firms that have been involved in financing discussions.
The proposal has not been publicly announced as a completed transaction. SpaceX, Nvidia and Apollo did not immediately respond to Reuters’ requests for comment, while PIMCO declined to comment. That distinction matters because the size of the financing reflects an ambition rather than a finalized spending commitment.
A $40 billion financing plan built around compute
The reported structure is striking because the money would not primarily be going toward rockets, launch infrastructure or spacecraft. It would be used to secure Nvidia computing hardware for the AI systems SpaceX is building and operating.
That fits with a broader shift already visible in SpaceX’s public statements. The company has been expanding its computing business alongside its traditional space operations, building large data-center clusters and selling computing capacity to AI companies.
On its first earnings call as a public company, SpaceX said it expected to finish 2026 with more than 2 gigawatts of compute online and suggested that its cumulative capacity by the end of 2027 could approach 10 gigawatts. The company also said it had decided to build its AI infrastructure exclusively around Nvidia technology, citing the Vera Rubin architecture.
That commitment is significant because it turns Nvidia hardware from one supplier among many into a strategic foundation for SpaceX’s expansion. It also helps explain why the company’s capital requirements are moving far beyond what would normally be associated with a launch provider.
SpaceX is building an AI business on top of its existing infrastructure
The shift is not simply about buying GPUs and putting them into conventional data centers. SpaceX has been developing a business around leasing compute capacity to AI companies, creating a potential source of recurring revenue from the same infrastructure it is spending billions to build.
Anthropic has publicly announced an agreement to use all of the compute capacity at SpaceX’s Colossus 1 data center, representing more than 300 megawatts and over 220,000 Nvidia GPUs. Reuters has separately reported that agreements between Anthropic and SpaceX could ultimately be worth as much as $84.5 billion through 2029, although those arrangements include cancellation provisions.
That model changes the financial calculation. Instead of treating AI hardware purely as an internal expense, SpaceX can attempt to deploy the equipment and sell access to the resulting computing capacity. SpaceX executives have argued that the economics of new compute deployments can produce relatively fast payback when demand remains strong.
It also explains why the company is willing to contemplate financing on a scale that would have seemed unusual for a rocket company. The underlying bet is that demand for AI compute will remain strong enough to support the debt used to build it.
Nvidia is becoming part of the financing story, not just the hardware supply chain
The SpaceX proposal arrives at a moment when the AI industry is experimenting with new ways to finance the enormous cost of computing infrastructure.
Nvidia announced in August that it was working with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR on independent financing platforms designed to mobilize more than $500 billion of third-party capital for AI infrastructure over time. Nvidia described compute infrastructure as an investable asset class, linking the economics of chips and data centers with long-duration financing.
That development provides important context for the SpaceX talks. The industry is no longer relying only on technology companies’ existing cash balances to fund AI expansion. Banks, private-credit firms, asset managers and bond investors are increasingly being pulled into the construction of the computing infrastructure required by large AI models.
In that environment, a company such as SpaceX can raise debt against a business that combines hardware, data-center capacity and contracted demand. The financing becomes easier to understand when viewed as infrastructure finance rather than simply a large technology purchase.
The timing matters because Nvidia capacity is becoming strategic
For AI infrastructure companies, having enough capital is only one part of the problem. They also need access to advanced processors, networking equipment, power and data-center capacity.
SpaceX’s decision to commit exclusively to Nvidia makes its expansion particularly sensitive to the availability and delivery schedule of Nvidia’s newest systems. Nvidia said in August that SpaceXAI would expand its AI infrastructure using the Vera Rubin platform and eventually adapt an optimized Vera Rubin system for orbital computing.
That is a much broader ambition than operating terrestrial AI clusters. SpaceXAI has described plans to extend accelerated computing into orbit through its planned Starmind AI satellite, using a modified Vera Rubin NVL72 system. The idea is to build a common computing architecture that can operate in large terrestrial AI factories and eventually in space.
The combination of terrestrial data centers, AI compute leasing and orbital computing helps explain the scale of the hardware requirements. It also shows why SpaceX’s AI strategy is increasingly intertwined with Nvidia’s product roadmap.
Why $40 billion is a consequential amount of debt
The financing would add another large layer of capital to a company that has already tapped public markets aggressively this year.
SpaceX completed an $86 billion initial public offering in June, according to Reuters, and subsequently announced a $25 billion inaugural investment-grade bond offering. SpaceX’s investor-relations site confirms the June bond issuance, which included multiple senior-note tranches.
A new $40 billion financing would therefore arrive on top of an already substantial capital program. The proposed structure also matters because investment-grade debt can attract institutional investors such as pension funds and insurers that generally have stricter limits on lower-rated corporate bonds.
The attraction for SpaceX is clear: debt can accelerate infrastructure deployment without requiring the company to finance every new data center entirely from equity. The risk is equally clear. AI infrastructure requires enormous upfront spending, while the revenue needed to justify that investment depends on sustained demand, high utilization and favorable pricing for compute.
That tension is already visible across the technology sector. AI companies and infrastructure providers are committing hundreds of billions of dollars to chips, power and data centers at a time when investors are becoming more selective about how quickly those investments can translate into cash flow.
SpaceX is making a much larger bet than a chip order
The most important part of the reported transaction may therefore not be the $40 billion figure itself. It is what SpaceX intends to build with the hardware.
The company’s strategy increasingly connects several businesses that were once easier to view separately: Starlink and other communications operations can provide a large existing revenue base; the launch business supplies space infrastructure; AI data centers create a new computing business; and future orbital computing could connect those systems even more closely.
SpaceX is also building toward its own chip manufacturing capacity. In August, the company announced Terafab, a planned semiconductor manufacturing facility in Texas that SpaceX says will combine logic, memory, packaging and testing. The company estimates the initial phase could require approximately $16.8 billion in capital investment from SpaceX and Tesla.
Terafab does not eliminate the need for Nvidia hardware in the near term. Instead, it illustrates the scale of the long-term strategy: SpaceX wants more control over the computing stack while continuing to deploy Nvidia’s leading AI architecture across its data centers and planned orbital systems.
The real question is whether compute demand can justify the capital
SpaceX’s proposed financing is ultimately a bet on the economics of AI infrastructure.
If demand for compute continues to outpace supply, large operators can potentially earn attractive returns by deploying processors early and selling access to customers that cannot build enough capacity themselves. SpaceX’s agreements with AI companies provide evidence that there is already significant interest in its capacity.
But the model has an important dependency: utilization. Billions of dollars in chips can produce strong economics when they are heavily used, but the same hardware becomes a major fixed cost if demand slows, customers cancel contracts or newer generations of accelerators reduce the value of older equipment faster than expected.
That makes the reported $40 billion financing more than a story about Nvidia’s popularity. It is a test of how far financial markets are prepared to support the next phase of the AI buildout, and how confidently investors believe the computing demand behind that buildout will persist.
For SpaceX, the direction is already clear. The company is no longer positioning AI as a side project around its space business. It is building an infrastructure business around compute, with Nvidia hardware at its center and increasingly large amounts of external capital behind it.
Sources: Reuters, Nvidia, SpaceX Investor Relations, Anthropic.
