Musk says SpaceX has completed $40 billion in GPU (chips used for AI computing) debt financing, with the related equipment expected to be delivered in Q1 and Q2 of 2027. On the early morning of October 26 Beijing time, Reuters reported the statement, while noting that the specific meaning of "completed" and whether investors have agreed to pricing remain unclear.

He attributed the financing capability to SpaceX's investment-grade status and the revenue generated by Starlink and its launch business, and predicted that new cloud service providers focused on AI computing power will face consolidation and bankruptcy. This gives a chip financing deal an implicit judgment on competitors' business models.

For investors, the question here is: even if demand for computing power continues to grow, how much profit can the companies providing it retain? Equipment must be purchased upfront, revenue is gradually realized only after delivery, and the financing costs in between are enough to change the value of an order.

However, diversified business revenue is not the only passport to financing. CoreWeave has already disclosed equipment financing that obtained an investment-grade rating. Musk's statement and peers' deals together point to one question: who do lenders actually believe will repay the money, and when will that money come back.

Before chips are delivered, financing costs are already affecting returns

Computing power expansion has a period during which money must be fronted. Purchasing equipment, preparing data centers, and deploying clusters all occur before continuous revenue collection. Even if customers are willing to buy computing power, operators still need to raise funds first in order to fulfill orders.

Therefore, the order amount cannot be directly equated with shareholder returns. Fulfillment requires equipment, power, operations, and financing costs; if revenue terms are already fixed, rising borrowing costs for new projects may compress the profit ultimately retained.

Delivery timing is equally important. After equipment arrives, it still needs to be deployed and accepted before it can form service capacity. Any delay may postpone cash collection, while whether debt has already begun accruing interest and when repayment starts depends on the specific contract—the two timelines may not be synchronized.

This is also the part worth questioning beyond the $40 billion: under what conditions can the funds be withdrawn, what is the cost, and when will they be repaid? Before these terms are clear, the scale of financing only indicates expansion intent and cannot alone prove how much cost advantage SpaceX has gained.

Starlink revenue is one form of support, customer contracts are another

Musk's explanation has its logic: if a company's revenue comes from multiple businesses, lenders may reference revenue sources beyond the computing project when assessing debt-servicing capacity, rather than relying entirely on AI returns that have not yet materialized.

But the group having revenue and a particular loan being able to obtain support from that revenue are two different things. The borrowing entity, the scope of guarantees, and restrictions on use of funds determine what creditors can actually recourse to. Starlink and the launch business cannot be directly understood as unconditional backstops.

Computing-focused companies, by contrast, can bring customer contracts to lenders. In addition to examining the operator, lenders also assess customers' payment capacity, contract terms, and whether equipment can be delivered as agreed. Customer credit can therefore become important support for project financing.

CoreWeave's financing disclosed in March is a reference: a maximum facility of $8.5 billion, with about $7.5 billion initially available, supported by related infrastructure and customer contracts, and the transaction received an investment-grade rating. The facility does not mean it has been fully drawn, and a project rating is not the same as the company's overall rating.

This transaction limits the explanation that "without non-AI revenue, you cannot borrow money." Diversified revenue and customer contracts can both support financing, but the sources of risk differ: the former requires testing whether the support can be transmitted, while the latter requires testing whether contracts can translate into stable collections.

Being able to borrow money may also leave renewal risk

Whether financing can be completed only answers where the money comes from. How long the loan tenor covers and how long customer commitments last determine what risks the project still needs to bear after expansion. Different transactions of the same company may also have completely different protection terms.

Another $2.6 billion loan disclosed by CoreWeave in August has a tenor of about five years, while the underlying customer contracts average about three years. This means that when part of the debt has not yet matured, the original customer contracts may already have ended, and the project needs to renew or find new customers.

The company believes this structure expands the range of customers it can serve and reflects lenders' willingness to bear renewal risk. But for investors, it also raises a question: after the old contracts end, can the equipment still be rented out at a sufficiently good price?

Therefore, the financing window cannot simply be divided into open or closed. Lenders can continue to provide funds through interest, collateral requirements, and repayment arrangements while increasing constraints on operations. Enterprises gain expansion capacity, but shareholders may not necessarily obtain a comparable improvement in returns.

$40 billion must pass the test of cash returns

In its October 20 analysis, Morgan Asset Management attributed recent pressure more to financing risk and sentiment changes rather than a clear deterioration in demand. This supports the judgment that credit conditions deserve attention and also reminds investors not to directly equate financing pressure with the disappearance of computing power demand.

For SpaceX, this statement still needs to be completed with specific terms: the actual borrowing cost, drawdown and repayment arrangements, and how the 2027 equipment delivery will connect with revenue. Only as this information gradually becomes clear can it be determined how much financing advantage the group's revenue actually brings.

For new cloud providers, the same test falls on whether customer contracts can continue to generate cash and cover equipment investment and debt repayment. Whether the shakeout predicted by Musk occurs will ultimately have to be proven by this period of cash returns, rather than decided by who announced a larger financing number.

Welcome to join the official BlockBeats community:

Telegram Subscription Group: https://t.me/theblockbeats

Telegram Discussion Group: https://t.me/BlockBeats_App

Official Twitter Account: https://twitter.com/BlockBeatsAsia