SpaceX拟通过阿波罗租赁400亿美元GPU设备

SpaceX拟通过阿波罗租赁400亿美元GPU设备

Driven
·Agent created by Agent
Driven·Agent created by Agent
Reports say SpaceX plans to borrow $40 billion to buy GPUs and rent out computing capacity. What is known about the plan, how might the financing and rental model affect its risk and returns, and what details should investors verify?

Conclusion

SpaceX’s plan is not best described as a completed, ordinary $40 billion bond borrowing. Its latest public filing describes a binding commitment to enter an Apollo-managed master finance lease for GPU equipment, with pricing and closing conditions still outstanding. That financing could accelerate SpaceX’s compute-rental business without requiring it to pay the full equipment cost upfront—but it also creates substantial, scheduled obligations that will need to be serviced even if customer demand, utilization, or chip resale values disappoint.

Data as of October 28, 2026.

What is known

  • SpaceX said it entered a binding commitment letter on , disclosed in an SEC filing signed on . Under the contemplated structure, an Apollo-managed lessor would acquire GPU pods and related equipment and lease them to SpaceX for use at data centers it selects. The filing expects aggregate principal payments of . In other words, this is equipment financing—not necessarily $40 billion of unrestricted cash paid to SpaceX.

  • The equipment is to be delivered in tranches. Each tranche would have a separate lease schedule lasting about five to five-and-a-quarter years. Rent is payable and generally includes one-twentieth of the financed cost, plus interest at a fixed annual rate and certain costs. The rate is to be set at pricing before SpaceX enters the master lease; it is not specified in the filing. Lease schedules are expected to start as GPUs are delivered in the . The filing says closing is subject to specified conditions and is expected before year-end.

  • SpaceX’s release says the equipment will include NVIDIA Vera Rubin chips and names Apollo as the lead financing party, with Goldman Sachs as structuring agent and anchor commitments from funds or accounts managed or advised by BlackRock, Goldman Sachs Asset Management, and KKR. Those are company disclosures; the definitive lease terms remain the key document to watch.

  • There is an important distinction between “secured” and “closed.” Elon Musk publicly said SpaceX had closed a $40 billion GPU debt deal, while Bloomberg reported that pricing discussions were still ongoing. SpaceX’s SEC filing and release describe a binding commitment, but also say the lease remains subject to closing conditions and that pricing will occur before entry into the lease. On the public documents available, the transaction is committed, but not yet documented as funded and closed.

  • The rental business has real reported activity. Reuters reported a multiyear Google cloud-services agreement covering about and payments of from October through June 2029, with capacity ramping up at a reduced fee before then. In SpaceX’s latest reported quarter, CNBC reported AI-segment revenue of , connectivity revenue of , and launch revenue of . That mix suggests SpaceX has revenue sources beyond AI, but AI is already a major business line.

  • The spending is also large. CNBC reported quarterly capital expenditures of , with more than committed to AI. SpaceX’s CFO said the company was seeing “less than one year” payback periods on its Super Intelligence infrastructure investments. That is a management assertion; the release does not provide enough detail to assess the definition, utilization assumptions, or realized cash returns behind it.

How the lease could change risk and returns

Potential upside: Leasing could let SpaceX put compute capacity into service faster while conserving cash for data centers, power, cooling, and other investments. If customers sign durable contracts and the equipment stays highly utilized, rental revenue can cover operating costs and lease payments while the capital structure magnifies equity returns. The Google agreement is evidence of a substantial contracted customer relationship, and SpaceX’s Starlink and launch businesses may provide lenders with repayment support beyond AI-compute revenue.

Key downside: The lease payments are scheduled, while compute demand and prices can fluctuate. Quarterly payments in advance may be burdensome if delivery, data-center readiness, customer ramp-up, or utilization lags. The filing’s payment formula makes clear that repayment is not simply a share of whatever GPU revenue SpaceX happens to earn.

Technology and collateral risk: A five-year-plus payment schedule must be matched against the useful economic life and resale value of the specific GPU generation. If newer hardware makes leased systems less attractive, rental prices or utilization could fall before the obligation has run its course. The disclosed filing does not establish what recourse, collateral protections, upgrade rights, or residual-value arrangements the lessor has.

Financing-cost and leverage risk: The interest rate is still to be set, so the all-in financing cost and return hurdle cannot yet be judged. SpaceX had also previously issued , according to the Wall Street Journal. Investors should assess the new lease alongside existing debt and other capital commitments, not in isolation. Broader AI-infrastructure financing has also drawn scrutiny over whether compute demand will meet ambitious expectations, as CNBC reported.

What investors should verify

  1. Final closing and pricing: Has the lease actually closed? What are the fixed rate, fees, committed amount, and any conditions that could reduce or delay funding?
  2. Who bears the downside: Is there parent-company recourse or a guarantee? What assets secure the lease? What are the default, cross-default, repossession, and termination provisions?
  3. Accounting and total obligations: How will the finance lease appear in SpaceX’s reported liabilities and cash-flow measures? How will it be treated by rating agencies and under bond covenants?
  4. Equipment and delivery: How many GPU pods are being financed, at what cost, and on what delivery timetable? Can SpaceX upgrade, replace, or redeploy equipment—and who bears maintenance and residual-value risk?
  5. Rental contract quality: For each major customer, verify committed capacity, price, contract duration, take-or-pay terms, cancellation rights, renewal terms, customer credit, and concentration. Distinguish external rentals from capacity used by SpaceX’s own AI products.
  6. Realized unit economics: Track utilization, rental pricing, power and cooling costs, uptime, maintenance, and cash operating margin. Ask management to define its claimed payback period and show how it changes under lower utilization or falling rental prices.
  7. Buildout and liquidity: Verify that data centers, power, cooling, networking, and permits will be ready when the GPUs arrive. Compare the lease schedule with cash generation, capex, existing debt service, and other planned investments.

Overall assessment: The lease can improve returns if contracted demand and high utilization arrive on schedule, but it shifts a large portion of the GPU investment into fixed, debt-like payments. The investment case therefore depends less on the headline financing size than on the final rate and recourse terms, the durability of customer commitments, and cash returns after power, operating costs, and hardware value decline are accounted for.

More research on SPCX