Elon Musk's SpaceX is moving deeper into the AI infrastructure race. According to a filing the company submitted to the US Securities and Exchange Commission (SEC) on Monday, on October 24 it signed a binding commitment letter for a lease of graphics processors (GPUs) and data center equipment with an entity managed by investment firm Apollo. The expected payments total about $40 billion.
How the deal works
According to the filing, the Apollo entity will buy the chips and related equipment and lease them to SpaceX for data centers the company chooses. According to reports, the chips are Nvidia's Vera Rubin, the company's next generation of AI chips. The equipment will be delivered in batches, and each lease will run about five to five and a quarter years, with rent paid quarterly in advance. The deal is expected to close by year-end, and the chips are expected to start arriving in the first half of 2027.
Debt or lease?
Musk wrote on X that "SpaceX closed $40B of debt for GPUs", adding that the company is investment grade and has Starlink and launch revenue. The official filing describes the deal as a finance lease, a long-term rental that works like a loan. It is also unclear exactly what Musk meant when he wrote that the deal had "closed", since according to reports the fixed interest rate has not yet been set.
Why SpaceX needs so many chips
Over the past year SpaceX has turned from a space and communications player into a company investing heavily in AI computing, and this deal reflects the scale of that investment. According to reports, Musk also warned that smaller cloud providers that rent out computing power, known as neoclouds, are likely to go through a wave of consolidation, because only players with strong balance sheets will have access to cheap financing.
Part of a wider trend
The deal joins a series of huge financing moves in recent days. Meta announced a plan to raise up to $95 billion in stock, and Broadcom arranged $63 billion in chip financing for Anthropic. The common thread is clear: even the richest companies in the world no longer fund the AI race from profits alone, but are turning to the capital markets and private credit funds.
For firms like Apollo, chips have become a new kind of asset, like aircraft or real estate, that can be bought and leased out for a steady return. The open question is what will happen to the value of these chips in five years, when newer and faster generations reach the market, and who will bear the risk if demand for computing power slows.
