Goldman, Apollo Shop $40 Billion SpaceX AI Chip Financing to Investors

Goldman Sachs and Apollo Global Management have started approaching investors to sell pieces of a roughly $40 billion financing deal backing SpaceX's AI chips, Bloomberg News reported on Wednesday. Some buyers are being asked to take stakes of up to $500 million before the formal syndication process begins, according to people familiar with the matter cited by the report.

It's a staggering sum for a single-purpose financing. The multi-tranche arrangement, structured to keep the debt off SpaceX's balance sheet, is the largest deal of its kind for GPUs on record, Bloomberg reported.

Reuters could not independently verify the report. Goldman Sachs declined to comment when asked. Apollo Global and SpaceX did not immediately respond to requests.

What is the deal actually funding?

The financing rests on an unusual structure: chips and servers as leased assets rather than purchased capital. SpaceX said in October that an affiliate of Apollo Global Management would buy AI chips and server hardware and lease them back to the company. The arrangement involves about $40 billion in principal payments.

The leases run a little over five years from delivery of the equipment. That delivery is expected in the first half of 2027, according to SpaceX's October statement.

The purpose is straightforward: financing SpaceX's growing computing needs as it expands its AI business. Rather than sinking billions into GPUs it must own and depreciate, SpaceX pays to use them, and the lessor holds the hardware as collateral.

Why syndicate a $40 billion deal?

Selling portions of the financing spreads the risk. No single balance sheet — not even Apollo's — wants to carry $40 billion of exposure to leased AI hardware on its own. By approaching investors ahead of formal syndication, Goldman and Apollo are testing appetite and price discovery for what amounts to a new asset class: securitized computing power.

The reported ask — stakes of up to $500 million per investor — suggests the firms are targeting large institutions, pension funds and asset managers capable of absorbing concentrated positions in a single-name, single-asset-type structure.

The stakes are high for the broader AI supply chain too. Neocloud operators and AI labs have scrambled for creative financing to acquire Nvidia-class GPUs, and a successful $40 billion syndication would establish a template others could copy. A rocky one would signal that investors see cracks in the AI hardware debt boom.

What comes next?

The timeline is now fixed on two dates: the informal sounding of investors underway this month, and the equipment delivery expected in the first half of 2027, when the five-year-plus lease terms kick in. How smoothly the syndication clears will shape whether GPU-backed financancing of this scale becomes a recurring feature of AI infrastructure funding or stays a one-off record.

(Reporting based on Bloomberg News and Reuters; Reuters could not independently verify the Bloomberg report.)