SpaceX CEO Elon Musk said on Nov. 26 Beijing time that the company has closed $40 billion in debt financing to buy the NVIDIA chips it needs for its AI business. In the same set of remarks, he delivered a verdict on the "neocloud" sector: a shakeout is coming, with large-scale consolidation and even failures among its players, and SpaceX will not be buying any of them.

The Money Is for Chips

In a post on X, Musk said the debt will mainly fund GPU purchases scheduled for delivery in the first and second quarters of next year, and that most of it is callable, so the impact on the current quarter is limited.

Two details are worth separating out. Delivery in the first and second quarters of next year means the computing power this money buys will only arrive, go online, and start generating revenue over the first half of 2027. None of it shows up on this year's books. Callable debt, meanwhile, hands the issuer an option: if rates fall or cash flow improves, it can retire the debt early rather than carry it to maturity. That is also why Musk says the impact on this quarter is small. The money is in, but the interest and principal burden has been pushed out.

An Investment-Grade Rating Is the Dividing Line

Musk credited two things for getting the financing done: the company's investment-grade credit rating, and the revenue coming from Starlink satellite internet and rocket launches.

Those two things are really one thing. A credit rating does not assess the story; it assesses where repayment comes from. Investment grade means rating agencies see a low probability of default, which translates into lower borrowing costs and easier subscription. Fall below investment grade and the same money costs far more in interest, or cannot be raised at all. What underpins the rating is Starlink and launch, two businesses that do not depend on AI. They generate cash consistently, regardless of how hot or cold the AI cycle runs.

Put another way, SpaceX is using the part of its business that is not betting on AI to backstop the part that is.

Neocloud Balance Sheets Were Never Built for This Market

Unlike the large cloud providers that serve customers across every industry, neocloud firms such as Nebius and CoreWeave sell mainly to technology customers, offering computing power purpose-built for AI workloads.

That positioning dictates their balance sheets. Their core asset is GPUs, and GPUs are bought with debt. The chips come first, the rental income later, with a stretch in between that is all spending and no revenue. The model moves fast when financing is loose: borrow, expand, and take on more orders. But the margin for error is thin. The moment the borrowing window shuts, expansion stops dead while the existing debt keeps accruing interest.

Musk's call lands squarely on that structure. For companies without meaningful revenue independent of AI data centers, he said, the debt markets have "almost completely closed" as a source of funding.

Once the Window Shuts, Only Two Exits Remain

If financing really does dry up, such companies have few options: get acquired by a larger rival, or fail.

Musk closed off the first one. He said SpaceX will not acquire any neocloud company, calling the idea "pointless." For a company that can already borrow, build, and buy chips on its own, purchasing another chip buyer adds nothing. What these firms lack is not computing power but the standing to raise cheap money, and that standing cannot be bought.

Markets have already voted with their feet. Investors had grown worried that the debt-driven AI expansion by neocloud and data center companies was unsustainable, triggering a selloff, and they have begun shorting such firms in the US.