SpaceX strikes $40 billion Apollo lease for Nvidia AI chips

By Grace Kim

SpaceX has struck a deal to lease specialized Nvidia graphics processing units (GPUs) for AI development from private equity manager Apollo Global Management for roughly $40 billion over about five years, plus interest and certain fees, the company said Monday.

SpaceX has not finalized the contract. It disclosed the terms in a filing with the US Securities and Exchange Commission, one day after Elon Musk previewed the GPU agreement on X.

Apollo will buy the processors and make them available to SpaceX. The hardware is Nvidia’s Vera Rubin GPUs, which the chipmaker began rolling out in the second half of this year. GPUs are considered the most powerful chips for building generative AI models.

What does the lease change for SpaceX?

The deal creates a five-year route to Nvidia hardware through a counterparty that purchases the chips and leases them to SpaceX. Under the structure, Apollo will make the processors available in phases, starting in the first half of 2027.

The $40 billion is the rough headline value; interest and certain fees will raise the total. The practical effect for SpaceX is therefore a future hardware and financing channel, not immediate capacity. The first scheduled releases fall in 2027, while Nvidia only began rolling out Rubin in the second half of this year.

SpaceX absorbed xAI in February. Before that merger, xAI built Colossus 1 and Colossus 2 in Memphis, Tennessee, in two years. Together, the two data centers house several hundred thousand GPUs.

Their servers and chips form cloud infrastructure, the computing capacity required to develop and run large AI models. SpaceX also rents that capacity to outside customers.

SpaceX is expanding Colossus 2. Musk said the site’s computing power should more than double by the end of the year compared with the level at the end of September. The added capacity would support SpaceX’s own model work and its business renting computing capacity to outside customers.

SpaceX rents capacity to Anthropic, the maker of the Claude chatbot, and to Google. Those contracts have become a major source of revenue, giving the company customers beyond its own AI operations.

Apollo’s existing financing role

Apollo has already helped finance Musk’s AI ambitions. In January, it led a $3.5 billion financing package that allowed an investment fund to buy Nvidia chips and lease them to xAI for Colossus 2.

Under the proposed agreement, Apollo would extend that role: instead of leading one financing package, it would buy the Vera Rubin processors and supply them progressively to SpaceX under the five-year lease.

Musk targets rival cloud providers

In his Sunday post on X, Musk said debt markets “have almost entirely shut off access to money for any company that doesn’t have substantial revenue independent of AI data centers.” He presented the Apollo agreement as evidence of investor confidence in SpaceX, contrasting it with cloud-market newcomers.

He described companies such as CoreWeave and Nebius as so-called “neoclouds” and wrote: “The 200+ neoclouds are going to see guppy-level death and consolidation. Some will make it through, but not many.”

Musk framed the financing gap as decisive: in his view, established AI data-center operators can attract capital while companies without substantial revenue outside AI data centers cannot.

For SpaceX investors, three checkpoints remain: contract finalization, phased Vera Rubin availability from the first half of 2027 and Colossus 2’s year-end capacity target. Until finalization, the $40 billion figure remains an estimate rather than a finalized contract value.