Skip to content
AI Usage3 min read

SpaceX Secures $40 Billion GPU Lease Deal With Apollo

SpaceX locks in a $40 billion financing deal with Apollo Global Management to lease Nvidia Vera Rubin GPUs and data center hardware for its AI infrastructure buildout.

AB

Author

AUG Bot

Published

Digital representation of GPU server racks and data center compute infrastructure financing

SpaceX Secures $40 Billion GPU Lease Deal With Apollo

Private equity structure finances Nvidia Vera Rubin chips for terrestrial AI infrastructure

SpaceX has entered into a binding agreement for approximately $40 billion in GPU financing with Apollo Global Management to fund its expanding AI compute infrastructure. Under the structure, Apollo will purchase Nvidia hardware and lease it back to SpaceX over five-year terms starting in early 2027.

Key details

According to regulatory filings submitted on October 26, 2026, Apollo Global Management will buy $40 billion worth of Nvidia next-generation Vera Rubin GPUs and specialized server hardware. Apollo will then lease the hardware back to SpaceX for terms lasting 5 to 5.25 years per equipment tranche.

The arrangement allows SpaceX to scale its compute footprint without drawing heavily on traditional debt markets. SpaceX reported capital expenditures of nearly $20 billion in Q3 2026—up from $6.9 billion in the same period last year—and expects Q4 capital spending to reach between $35 billion and $40 billion. Company executives noted that its AI segment, which powers Grok, Cursor, and cloud compute hosting, generated $7.6 billion of its $14.6 billion total Q3 revenue.

Why this matters

The $40 billion deal highlights the staggering capital intensity required to acquire frontier AI compute hardware. By utilizing sale-and-leaseback structures, major AI operators are creating specialized financing mechanisms to acquire hundreds of thousands of high-density GPUs while avoiding direct balance sheet leverage.

Context

As hyperscalers and neoclouds compete for GPU capacity, traditional debt financing has become tighter for entities without diversified revenue streams. SpaceX's cash flow from Starlink and launch operations enabled it to secure investment-grade terms, reflecting a widening divide between cash-generative conglomerates and pure-play AI compute providers.

What happens next

The transaction is expected to close by the end of 2026, with initial GPU equipment deliveries and lease terms commencing in the first half of 2027. SpaceX plans to deploy the Vera Rubin platform across its expanding data center facilities to support growing compute demands for model training and cloud inference.


Source: TechCrunch Published on AI Usage Global, author: AUG Bot

Related

Read more

More posts that expand on the topics, companies, and AI trends covered in this story.

Digital representation of AI data center power orchestration and server rack energy management
AI Usage

Karman Boosts NexGen AI Server Capacity 60% in Montreal

Karman deploys AI power orchestration software at NexGen Cloud's Montreal data center, allowing 16 Nvidia B300 servers to run within a 10-server power budget.

Anthropic Claude branding and data center compute infrastructure
AI Usage

Anthropic Discloses $417B Compute Commitments in S-1 Filing

Anthropic's SEC S-1 IPO filing reveals $417 billion in long-term compute and data center payment obligations, highlighting the immense infrastructure costs of frontier AI scaling.

Aerial view of data center land development in Missouri
AI Usage

AWS Plans 15-Building Data Center Expansion in Missouri

AWS proposes a 15-building expansion across 890 acres in Montgomery County, Missouri, alongside community talks for potential data center development in Posey County, Indiana.