SpaceX is negotiating a $40 billion debt package, led by Apollo, with a single purpose: buying Nvidia chips for its AI data centers. This is not just another purchase; it is an X-ray of how AI infrastructure is being financed.
According to the Financial Times and Bloomberg, SpaceX is in talks to borrow $40 billion. The package would combine roughly $10 billion in bank loans with $30 billion in investment-grade debt. Apollo Global Management is leading the deal and would resell the debt to a broad base of investors; bond fund Pimco is among the lenders reportedly in talks.
There are important caveats. Bloomberg describes the talks as "early stage" and warns they could end without a deal; the FT says closing would be expected in 2027. None of the companies involved has publicly confirmed the plan: everything comes from anonymous sources.
TIKR’s analysis points out that Nvidia appears on three sides of the deal at once. It is the chip supplier; according to that source, it is also a SpaceX shareholder (about 122.8 million shares, around $21 billion as of August 2026); and in August Apollo signed preliminary agreements to create Nvidia compute-financing platforms backed by more than $500 billion in third-party capital. In August, Elon Musk said SpaceX would build its data centers using only Nvidia hardware.
The paradox the same source notes: SpaceX would hold $93.5 billion in cash and $6.5 billion in short-term investments, and yet it is seeking outside debt, preserving liquidity while its capital spending runs at around $74 billion a year. For Nvidia, the $40 billion would equal roughly one fifth of its annual data center revenue ($193.7 billion in fiscal 2026, according to TIKR). Those figures come from a single analysis, so treat them as an order of magnitude, not audited data.
Three models coexist today: (1) own cash, as the big hyperscalers do; (2) debt backed by the hardware or by compute lease contracts, which is what this case looks like; and (3) cross-guarantees between suppliers and customers, such as Broadcom’s backing of Anthropic’s financing, which we already covered on the blog. The key difference is where the risk lands: in the debt model, if compute demand slows, the lender loses and the buyer is left with depreciating chips.
TIKR cites a Morgan Stanley estimate that AI infrastructure will need $1.5 trillion in external financing by 2028. It is a bank’s projection, not a fact, but it explains why these packages are becoming common.
Sources: Financial Times and Bloomberg (via The Next Web and InvestingLive), TIKR, Forbes Italia and HotHardware. The figures are reports from anonymous sources and have not been confirmed by the companies.
Carlos Montiel is an enterprise AI solutions architect. He implements LLMs, Agents, RAG, and orchestrators for companies across Guatemala and Latin America. Reach out for a consultation.
Contact Carlos Montiel