SpaceX commits to $40B Nvidia GPU lease via Apollo

By VK

Published 27 Oct 2026, 17:48 UTC

Reporting from TechCrunch

SpaceX signed a binding commitment for a roughly $40 billion finance lease of Nvidia Vera Rubin GPUs from an Apollo-managed entity.

What it means for founders

Diversified balance sheets may get the money. Musk argues lenders now favor borrowers with revenue outside AI; if he is right, the gap between giant labs and smaller compute buyers could widen.
Check how your GPU cloud is funded. Neoclouds lean on asset-backed debt too, as with Lambda's GPU-secured loan and planned equity raise. Tighter credit could change their pricing and capacity, so keep workloads portable.
New Grok capacity is a 2027 story. Deliveries are expected in the first half of 2027, so this compute is not imminent.

The story

SpaceX has signed a binding commitment to enter a master finance lease for Nvidia GPUs and supporting gear, valued at about $40 billion, with an Apollo Global Management vehicle as lessor. The company disclosed the arrangement in a Form 8-K filed Monday, October 26, 2026, and Apollo published SpaceX's announcement the same day. SpaceX calls it the largest committed GPU financing.

How the lease works

Per the filing, the Apollo-managed lessor will buy GPU pods and related equipment and rent them to SpaceX for data centers SpaceX picks. Hardware will arrive in tranches, each on its own lease schedule lasting roughly 5 to 5.25 years.

Rent is due quarterly in advance and generally equals one-twentieth of the tranche's financed cost, plus a fixed-rate charge on whatever balance is still outstanding, plus certain ongoing costs. That rate will be set at pricing, before the master lease is signed. SpaceX expects principal payments across all schedules to total about $40 billion.

Per the release, Apollo and Goldman Sachs lead the financing parties, with Goldman the sole structuring agent. Anchor commitments came from funds and accounts run by BlackRock and Goldman Sachs Asset Management, and from KKR-advised capital accounts. The chips are Nvidia's Vera Rubin generation, slated for what SpaceX calls its terrestrial compute buildout. The deal should close before 2026 ends, and each lease schedule should begin as its hardware arrives during the first six months of 2027.

Why SpaceX wants it

CFO Bret Johnsen says the lease broadens the company's funding mix and addresses a notable share of the money its growth plans require. He also says SpaceX is seeing payback periods of under a year on its infrastructure investments as it adds compute for Grok and cloud clients.

TechCrunch reports that third-quarter capital expenditures neared $20 billion, against $6.9 billion a year earlier, and that Johnsen guided to $35 billion to $40 billion for the fourth quarter. SpaceX's AI segment, which TechCrunch says spans Grok, the Cursor coding tool and cloud compute, produced $7.6 billion of $14.6 billion in quarterly revenue.

A tense credit backdrop

CNBC reported last week that CoreWeave, Nebius and IREN shares had each fallen well over 30% in October, and that CoreWeave's credit default swap spreads had passed 11%. Musk, citing SpaceX's Starlink and launch cash flow, argued on X that GPU-rental specialists face a shakeout, writing that "the 200+ neoclouds are going to see guppy-level death and consolidation," per TechCrunch.

PitchBook notes the lease is bigger than a $36 billion notes package from this summer, led by Apollo and Blackstone, that buys and leases GPUs to Anthropic. The commitment also sits awkwardly beside Apollo chief economist Torsten Slok's recent note that AI shows no payoff outside tech.

What we don't know yet

The fixed rate, which is set at pricing.
How many GPUs $40 billion buys, and which data centers will host them.
Whether closing conditions are met on time; the filing warns the deal may not close as expected.