Rothschild Bearish on New AI Cloud Giants, "The Big Short" Warns of $3 Trillion Off-Balance-Sheet Commitments

The prosperity narrative of AI computing power infrastructure is now facing systemic challenges from the credit market.

The independent research arm of Rothschild, Rothschild & Co. Redburn, has for the first time initiated coverage on CoreWeave and Nebius, assigning both a "Sell" rating. The core judgment: The stock market continues to price in AI growth aggressively, while the credit market has begun to price in risks that the equity market has largely ignored.

At the same time, Michael Burry, known as the "Big Short," estimates that top technology giants together hold more than $3 trillion in off-balance-sheet commitments, raising concerns externally about the sustainability of their expansion capabilities.

In a report released Monday, Rothschild & Co. Redburn stated that new cloud enterprises exhibit "cyclically inflated" earnings and valuations, with Nebius' stock price already reflecting the most optimistic outcomes amid ongoing revaluation of compute economics and facing the greatest downside risk; while the risk-reward ratio for CoreWeave "remains insufficiently attractive."

The institution also noted that, once hyperscale cloud providers' off-balance-sheet commitments are taken into account, their actual economic leverage is higher, potentially limiting their ability to fund the next wave of large-scale AI infrastructure construction.

These doubts did not immediately suppress market sentiment. On Monday, Nebius closed up 4.14%, CoreWeave closed up 5%; retail investor sentiment towards both remains in the "bullish" zone, significantly diverging from the cautious stance of research institutions.

Rothschild’s First Coverage: Dual "Sell" Ratings for New Cloud Leaders

Rothschild & Co. Redburn’s coverage includes several key companies in the AI computing infrastructure sector.

For CoreWeave and Nebius, the institution issued "Sell" ratings; for six companies transitioning from bitcoin mining to AI computing power infrastructure—Core Scientific, TeraWulf, Applied Digital, Cipher Mining, Hut 8, and IREN—it issued "Neutral" ratings across the board.

The ratings differentiation for the two types of companies reflects differing assessments of business model risks.

New cloud companies directly engaged in GPU leasing bear higher equipment depreciation, financing costs, and order fulfillment pressures; in contrast, mining firms transitioning via in-house power resources and data center capacity have more flexible entry paths into AI infrastructure and relatively fewer valuation concerns.

Rothschild also pointed out that traditional real estate investment trusts, "as indirect beneficiaries that have not yet received sufficient attention, are especially noteworthy."

Price Increase Logic Supports Stock Prices, but Financing Costs Erode Profits

The shared logic supporting CoreWeave and Nebius’ stock prices is the persistent expansion of AI training and inference demand, while scarce, immediately deliverable GPU clusters, power, and data center capacity remain limited.

On September 17, Nebius announced about a 20% price hike for some on-demand GPU rentals effective October 1—the second price increase in three months; CoreWeave also stated that it is signing new contracts at higher prices. Nebius's stock price has risen approximately 167% year-to-date.

However, Rothschild’s key question targets profit quality: How much return can GPU rental income ultimately leave for shareholders after covering hardware depreciation, operational expenses, and financing costs?

CoreWeave's Q2 revenue reached $2.575 billion, with net interest expenses for the same period at $640 million; financing costs are now a key variable affecting profitability.

The firm believes that, in the AI inference era, the new cloud competition will focus more on order quality, prepayment arrangements, equipment utilization, and capital returns after covering hardware upgrades and financing costs.

The Credit Market Sends a Starkly Different Signal From the Stock Market

Rothschild & Co. Redburn points out that the credit market is "sending a more significant signal." The credit market includes tools such as corporate bonds, bank loans, private credit, data center project financing, and credit default swaps (CDS).

While equity investors focus on AI growth potential, creditors pay closer attention to whether future cash flows cover interest and principal. Widening credit spreads and rising CDS costs usually indicate that investors are demanding higher risk compensation.

The specific example cited by the institution: Oracle's Project Jupiter data center’s $18 billion debt, trading at about 89 to 91 cents on the dollar face value, reflecting investor caution about project progression and financing risk.

According to the institution’s logic, including off-balance-sheet commitments, hyperscale cloud providers’ real economic leverage is even higher. If their expansion capacity becomes constrained by tighter financing conditions, pressure would be transmitted to CoreWeave and Nebius’s growth and valuations via slower new orders, renewal negotiations, and counterparty credit risk.

Burry Estimates Over $3 Trillion in AI Company Commitments, Exposing Capital Recovery and Hardware Upgrade Mismatch

Rothschild’s credit market warning echoes the latest views from famed short-seller Michael Burry.

According to reports, Burry estimates that Amazon, Meta, Alphabet, Microsoft, and Oracle collectively hold nearly $1.2 trillion in uninitiated lease commitments, over $1.5 trillion in purchase obligations, and together with special purpose entities, guarantees, and other contingent exposures, total more than $3 trillion.

Burry made it clear that he is not alleging fraud; his primary concern is that long-term leases and procurement expenditures are already locked in, yet equipment’s economic value, technological requirements, and future demand may change even faster—exposing a structural mismatch between capital recovery and hardware updates for companies.

Notably, this $3 trillion is a cumulative estimate of various long-term commitments, not directly equating to actual losses; investment implications mainly lie in the fact that assessing the expansion abilities of AI data center giants also requires close attention to future cash outflow obligations disclosed in financial statement footnotes.

For investors, Rothschild's report does not mean order cancellations or company guidance downgrades, but may prompt some institutional investors to assess order growth alongside capital recovery capabilities, while re-examining differences in return structures among compute operators, colocation data center providers, and power resource owners.