SpaceX Closes $40 Billion GPU Financing Deal With Apollo for AI Data Center Buildout

SpaceX has closed a $40 billion equipment financing deal with Apollo Global Management to purchase Nvidia GPU pods for AI data centers — the largest GPU financing commitment ever announced. The deal makes SpaceX a major player in the AI compute buildout, alongside established cloud hyperscalers, and further strains an already constrained Nvidia supply chain.

SpaceX CEO Elon Musk confirmed the deal closed on X, describing it as the largest GPU financing arrangement to date. According to Bloomberg, deliveries of the GPU pods will begin in the first half of 2027 under five-year lease schedules. Apollo Global Management, a private equity and credit firm managing over $700 billion in assets, is providing the debt financing — essentially loaning SpaceX the capital to buy the hardware, with lease payments spread over five years.

What SpaceX Is Building

SpaceX has been steadily building an AI infrastructure arm, separate from its launch and Starlink satellite internet businesses. The company operates Grok, the AI assistant developed by xAI (Musk's separate AI company), and has been expanding its compute footprint. This deal positions SpaceX as a hyperscaler — a company that operates AI compute at the scale of Amazon Web Services, Microsoft Azure, or Google Cloud — though the specifics of which customers or workloads will run on this hardware have not been disclosed.

Key deal parameters:

• Total value: $40 billion
• Financier: Apollo Global Management
• Hardware: Nvidia GPU pods
• Delivery window: H1 2027
• Lease term: 5 years
• Confirmed by: Elon Musk on X

The Capital Arms Race in AI

This deal is the latest data point in a capital arms race that has no precedent in the tech industry's history. Microsoft has exceeded $50 billion in quarterly capex. Google and Amazon are each spending tens of billions annually on AI infrastructure. Sovereign wealth funds from the UAE, Saudi Arabia, and Japan have announced multi-hundred-billion-dollar AI investment programs.

The constraint throughout has been Nvidia GPU supply. Nvidia's H100 and its successors are manufactured by TSMC in Taiwan, and production capacity cannot be spun up quickly. A commitment of this size by SpaceX — locking in hardware deliveries 6-9 months out — signals that large organizations are willing to finance GPU inventory long before they have customers lined up.

For businesses and developers, the long-term effect of this buildout is deflationary: more compute capacity means lower inference prices over time. OpenAI's 50% price cut for GPT-6.1 Terra — announced the same day — is already reflecting this trajectory.

What to Watch

Watch whether other non-traditional tech companies (aerospace, defense, industrial) follow SpaceX into equipment financing arrangements to secure GPU supply. Also watch how Apollo and similar credit firms structure these deals — GPU depreciation curves are steep, and the risk calculus on five-year leases for AI hardware is genuinely novel.

Sources: Bloomberg