In brief: SpaceX says an Apollo-managed affiliate will acquire AI chips and server equipment and lease it back to the company. Principal payments across the arrangement total about US$40 billion, turning accelerators into long-lived financed infrastructure rather than an ordinary technology purchase.
The scale of AI computing is changing how companies pay for hardware. Instead of buying every GPU upfront, SpaceX has arranged for a financing vehicle to own equipment and rent it to the company over multiyear terms.
Reuters reported on 26 October 2026 that an Apollo Global Management affiliate will buy AI chips and server hardware and lease them to SpaceX. The transaction carries about $40 billion in total principal payments. Each lease is expected to run a little over five years from the equipment’s arrival.
The headline number is a payment commitment
The $40 billion figure should not be read as a cash payment made today, the value of computing already installed, or revenue earned from AI customers. It is the principal associated with leases expected to begin after hardware delivery, currently planned for the first half of 2027. SpaceX aims to finalize the arrangement by year-end.
In simplified terms, the financing affiliate pays for the servers and chips. SpaceX gains the right to use them and makes scheduled payments. Depending on final accounting and contract terms, such structures can match cash outflows to the period in which equipment generates revenue while giving lenders claims linked to hard assets and contracted payments.
Why finance chips like aircraft or power plants
AI accelerators are expensive, revenue-producing assets installed in facilities with dedicated power, cooling and networking. That makes them increasingly suitable for infrastructure-style finance. The owner can underwrite the credit of the lessee, the residual value of the hardware and the predictability of lease payments.
Apollo has used similar structures before. In January 2026 it announced a $5.4 billion financing involving Valor and AI compute, describing the equipment as collateral in a five-year facility. The new SpaceX agreement is much larger and underscores the expansion of private credit into data-centre hardware.
Leasing changes risk; it does not remove it
Hardware can lose value quickly when a new accelerator generation offers better performance per watt. A five-year contract therefore spans several product cycles. Lease economics must absorb utilization risk, technology obsolescence, maintenance and the possibility that secondary-market values fall faster than expected.
SpaceX also needs facilities, electricity and cooling before delivered chips can produce useful compute. Financing secures a route to hardware; it does not prove that every system will arrive on time, operate at high utilization or generate returns above the cost of capital.
The structure can nevertheless protect liquidity compared with an immediate cash purchase. SpaceX has revenue from launch and connectivity businesses as well as its AI operations, giving lenders more than one activity to assess. That diversification may distinguish it from specialized “neocloud” providers whose income depends mainly on renting compute.
A new layer in the AI supply chain
The AI boom is usually described through chips, models and data centres. This deal makes a fourth layer visible: balance sheets. Accelerators must be financed before they can be installed, and the cost and availability of credit can constrain capacity as surely as fabrication or grid connections.
The announcement is a commercial financing agreement, not a performance benchmark or a forecast of guaranteed AI profits. Its importance is that a $40 billion equipment lease now looks plausible for one company. AI infrastructure is becoming large enough to borrow the financial machinery used for fleets, factories and power projects—and to inherit their debt risks.
This article explains an infrastructure transaction and is not investment advice.
Leave a Reply