Elon Musk’s SpaceX is in talks with banks and investors to raise $40bn of debt to buy chips from Nvidia, according to people familiar with the matter, in what would be one of the largest single financings yet assembled to buy computing hardware.
The company is seeking about $10bn in bank loans and $30bn in investment-grade debt to pay for the chips, the Financial Times reported, citing unnamed sources. Apollo Global Management is leading the financing, and the deal is expected to close in 2027.
Pacific Investment Management is among the investment managers looking at the deal, the sources said, speaking on condition of anonymity because the talks are private. The discussions are at an early stage and could end without a deal being completed.
Nvidia, SpaceX and Apollo did not immediately respond to requests for comment on the FT’s report.
The scale is not unusual for the sector. Technology companies and AI model developers are investing and borrowing hundreds of billions of dollars to build and lease giant data centres full of chips, and the worldwide expansion of computing capacity has sent the costs of everything from land to chips to power-generation equipment soaring, touching off a series of very large financing deals.
Musk has given figures for his own build-out. Colossus 2, an AI computing cluster run by his xAI business in the Memphis area, now has 110,000 Nvidia GB200 chips and 440,000 GB300s, he said in a post on X. A further 220,000 GB300s are due to be operational next week and another 220,000 in November, with an additional 220,000 possibly coming online in late December "if we get lucky".
SpaceX also aims to launch orbital AI computing tests by the end of 2027.
The lending side was organised in advance. In August, Nvidia tapped a group of US investment firms for a high-profile partnership to source $500bn in financing for AI infrastructure. The coalition — Apollo, Blackstone, BlackRock, Brookfield Asset Management, Goldman Sachs Group and KKR & Co — said at the time that it would "create dedicated pools of capital at significant scale at attractive rates for Nvidia customers", with executives indicating a focus on debt financing to provide access to compute for Nvidia’s largest customers.
Set the two halves of that side by side and the structure becomes clear, and rather elegant. Dedicated pools of capital were created for Nvidia customers. A firm from those pools is leading a $40bn raise. The $40bn is to be spent on Nvidia chips. Demand, at this scale, is a financing structure.
This resolves a question that has troubled the sector all year, which is how to distinguish demand from supply. The answer is that there is no need to, because the same money performs both roles in sequence, and does so at attractive rates.
Our analysts have measured the thing properly. Ten billion dollars of bank loans plus $30bn of investment-grade debt equals $40bn of chips, which equals $40bn of chip revenue, against a stated commitment to source $500bn. At that ratio the coalition can underwrite the demand for its own product twelve and a half times over before anybody has to want anything.
The Department of Unnecessary Research notes that the financing is expected to close in 2027 and the orbital computing tests are expected by the end of 2027, so the chips may be ordered, financed, delivered and placed in orbit inside the same accounting year as the loan used to buy them. It therefore recommends the one remaining efficiency available: send the capital directly, and leave the chips where they are.