Technology

SpaceX Chip Lenders Ask That Financed Hardware Stay Out of Orbit

Martin HollowayPublished 2m ago4 min readBased on 1 source
Reading level
SpaceX Chip Lenders Ask That Financed Hardware Stay Out of Orbit
Photo by NASA Goddard / Public domain

Lenders behind SpaceX's $40 billion chip financing have asked that none of the purchased chips be used for the company's orbital data center plans, the Financial Times reported. The report, cited by TechCrunch on November 20, 2026, places a use restriction at the center of a debt structure that is normally built on the opposite assumption: that the hardware can be moved, resold and recovered.

Data center financing commonly uses the accelerators themselves, the GPUs and similar chips built for AI workloads, as collateral. That is the asset a lender can claim if the borrower defaults. The lender can repossess the chips and resell them, but that remedy only has value if the hardware keeps its worth outside the facility where it was installed. An orbital deployment would put that premise under pressure. A lender's recovery depends on being able to physically reach and sell the asset, which is a different question from whether the chip works at all.

Residual values are already moving unevenly. TechCrunch reports that some used H100 accelerators are selling for about half their original 2023 cost basis, the price at which they were booked as an asset. Some H200 configurations, by contrast, are selling for the same price they carried when introduced in 2024. Two chips from adjacent generations, both in the secondary market, are following different depreciation curves. Anyone underwriting accelerator collateral, meaning assessing the risk before extending a loan, has to model that variance by configuration rather than by product family.

On the space side, Elon Musk said beginning in early 2026 that space will be the cheapest place to operate data centers in two to three years. That is a forecast about operating economics, and it is the kind of claim that the financing structure must eventually reconcile with. Google, for its part, is deploying its own TPU chips, its custom tensor processing units for AI, in space through Project Suncatcher. The two efforts point the same direction, toward compute placed outside terrestrial facilities, but they differ in how much of the economics each has disclosed.

In my view, the lender restriction is the more concrete signal in this story. Musk's cost projection and Google's program describe where the industry hopes to go. The request from the lenders describes what the people holding the debt are willing to accept today, and it suggests that the collateral model has not yet caught up with the deployment model. A chip in orbit is not one a lender can walk into a colocation hall and remove, so the recovery assumptions behind accelerator-backed debt would need a different foundation.

The broader context here is that the space environment raises its own engineering questions. Chips must tolerate radiation, which can flip bits or degrade circuits, and they must survive thermal cycling, the repeated heating and cooling of orbit. Those questions are separate from the financing questions, but they are linked. Lenders price risk on expected service life and recoverable value, and both depend on how the hardware behaves after launch.

The optimistic reading is that these constraints are solvable and that they are being worked on in public. If orbital compute proves economical, it could widen the set of places where data processing can happen, and the financing industry has adapted before to new classes of hardware. For now, the reported restriction marks the boundary between the stated ambition and the terms lenders are currently willing to sign.