SpaceX commits $40 billion to GPU financing with Apollo

By Elena Vasquez

SpaceX has committed roughly $40 billion to finance AI chips and related data center equipment, the company disclosed in a regulatory filing Monday. The binding commitment arrives as SpaceX races to expand the compute behind Grok and its cloud services.

SpaceX called the deal the largest committed GPU financing. It gives the company another route to fund that expansion without relying solely on traditional debt markets.

An entity managed by Apollo Global Management will buy the chips and related server hardware, then lease the equipment to SpaceX. The leases will last about five to five-and-a-quarter years, beginning when each tranche of equipment is delivered.

SpaceX expects the deal to close by year-end, subject to closing conditions. The leases will start in the first half of 2027 as the GPUs and related equipment arrive.

CFO Bret Johnsen said in a Monday statement: “This lease diversifies our funding sources and addresses a notable portion of the funding required to deliver on our growth plans.” He also said SpaceX is seeing less-than-one-year payback periods on its AI infrastructure investments as it scales compute.

The company is expanding this infrastructure alongside Starlink and rocket launches. The financing therefore supports a broader buildout rather than a stand-alone chip purchase.

What does Apollo finance?

Apollo Global Management is the New York-based private equity giant managing the buyer. Its entity will own the chips and servers, while SpaceX leases the equipment for its terrestrial compute buildout.

Apollo said the financing will help deliver Nvidia’s next-generation Vera Rubin chips for SpaceX’s terrestrial compute buildout. The reference to a named chip generation ties the lease to a specific hardware roadmap.

Ownership will sit with the Apollo-managed entity; use will sit with SpaceX under each lease. That separation gives SpaceX another funding option beyond traditional debt markets, according to the filing.

The leases will not all begin on the same day. Each starts with the delivery of its equipment tranche, and the roughly five-year terms run from those separate dates.

SpaceX expects the financing to close by year-end if it satisfies the stated conditions. The close comes before the first lease. Once those conditions are met, SpaceX expects the first leases in the first half of 2027.

The hardware will arrive in stages, so the leases will also begin in stages. Each five-year commitment stays tied to the equipment it finances as SpaceX expands its terrestrial computing capacity.

How large is SpaceX’s AI business?

SpaceX’s AI segment includes the Grok chatbot, the Cursor coding tool and its cloud compute business. The segment outperformed investor expectations in the third quarter.

It brought in $7.6 billion of SpaceX’s $14.6 billion in revenue. That amount exceeded what Starlink and rocket launches produced together.

The segment groups three services under one reporting line. Its $7.6 billion therefore reflects more than the Grok chatbot; the figure also covers Cursor and the company’s cloud compute operation.

Capital expenditures reached nearly $20 billion, up from $6.9 billion a year earlier. SpaceX said the spending surge came as it raced to expand its compute footprint.

On SpaceX’s Oct. 20 earnings call, Johnsen said the company expects to spend $35 billion to $40 billion in the fourth quarter alone. That forecast extends the sharp spending increase into the next quarter.

The quarter paired strong segment revenue with a sharp rise in capital spending. That combination makes financing central to how quickly SpaceX can add compute capacity for Grok, Cursor and its cloud services.

In the third quarter, AI supplied more revenue than Starlink and rocket launches combined. SpaceX can therefore make a case for financing future growth around a business that already produces substantial revenue.

A GPU-rental specialist has a narrower case. Its hardware remains central to the business, but it lacks the same range of revenue sources outside AI infrastructure.

Why use a lease now?

Debt markets have grown turbulent for AI infrastructure players. Investors have become squeamish about runaway capital expenditure and limited profits.

SpaceX’s own access to capital remained strong. The $40 billion commitment showed that the company could still tap Wall Street, while Musk used the announcement to take a victory lap over pure-play AI compute rivals.

Musk pointed to SpaceX’s investment-grade status and steady cash flow from Starlink and rocket launches. He argued that debt markets have largely closed their doors on companies without major revenue outside AI data centers.

More than 200 neoclouds were built solely to rent out GPUs for AI workloads. In Musk’s framing, those companies lack the same mix of non-data center revenue that SpaceX has.

“That means the 200+ neoclouds are going to see guppy-level death and consolidation,” Musk wrote on X. “Some will make it through, but not many.”

The dispute turns financing access into a competitive issue. Musk’s argument is less about GPU technology than about which operators can finance large fleets of chips and absorb investor scrutiny.

The backdrop matters because the market is already funding an expansion on a huge scale. SpaceX plans $35 billion to $40 billion of fourth-quarter spending, while more than 200 GPU-rental specialists seek capital.

That combination of heavy spending and investor caution puts pressure on companies with weaker cash flow. Musk expects consolidation among the neoclouds, even if some survive.

The Apollo structure gives SpaceX access to hardware through leases rather than requiring one immediate purchase in full. That option matters against the $35 billion-to-$40 billion fourth-quarter spending plan.

When will the GPUs arrive?

Closing and lease commencement are separate milestones. SpaceX expects the financing to close by year-end and the first leases to begin in the first half of 2027.

The first leases will start as GPUs and related equipment arrive. Apollo said the arrangement will help deliver Nvidia’s Vera Rubin chips for SpaceX’s terrestrial compute buildout.

The schedule links the financing directly to SpaceX’s wider infrastructure expansion. Starlink and rocket launches remain part of the business mix, while AI infrastructure spending has surged alongside them.

By the time deliveries begin, the spending surge will already be underway. Capital expenditures reached nearly $20 billion in the third quarter, and Johnsen expects $35 billion to $40 billion in the fourth quarter.

Musk’s claim puts the financing race in direct terms: companies with other cash flows may keep building, while GPU-only operators face consolidation. SpaceX’s own agreement supplies the example he used to make that case.

If SpaceX closes the deal by year-end, hardware-backed leases start in the first half of 2027. For the 200-plus GPU-rental rivals Musk targeted, access to long-term capital will shape which companies keep building and which consolidate.