AI Computing Power Boom Faces Doubts from Influencers: Rothschild Shorting "AI New Cloud," "Big Bear" Warns of $3 Trillion Off-Balance-Sheet Commitments

According to The Smart Investor, some bullish Wall Street analysts believe that new cloud AI leaders such as Nebius (NBIS.US) and CoreWeave (CRWV.US) will have broader and more sustainable business opportunities in the era of AI inference. However, Rothschild & Co. Redburn, an independent investment research institution under the Rothschild Group, conveyed a clear sense of caution and a selling logic in its first coverage of these two new cloud leaders, stating that the credit market is issuing warning signals. The institution believes that the profits and valuations of new cloud companies are currently being inflated by the boom cycle, particularly Nebius, whose share price has already reflected the most optimistic scenario regarding the re-rating of compute economics. CoreWeave has relatively less downside risk, but the risk-reward profile is still insufficient to support a positive rating.

The shared logic supporting the relatively strong performance of CoreWeave and Nebius stocks this year lies in the expansion of AI training and inference demand, while GPU clusters, electricity, and data center capacity that can be delivered on time remain scarce. This scarcity has prompted some bullish Wall Street analysts to raise expectations for revenue scale and pricing power for these two new cloud companies. In particular, Nebius' share price has risen about 167% year to date.

On September 17, Nebius announced it would raise some on-demand GPU rental prices again from October 1, with reports summarizing the increase at about 20%. This is the second price hike in three months. CoreWeave also stated it is signing contracts at higher prices. These changes strengthen the bullish logic that “compute demand can be monetized,” particularly as Nebius' latest adjustment targets some on-demand services while existing long-term contracts continue under their original terms.

From the perspective of data center project economics, the Rothschild report probes: After covering equipment depreciation, operating expenses, and financing costs, how much is left as shareholder return from GPU rental income? CoreWeave’s revenue in the second quarter reached $2.575 billion, with net interest expenses during the same period at $640 million, indicating that financing costs are already a major factor affecting profitability. Rothschild further argues that off-balance sheet commitments by hyperscale cloud providers amplify their actual economic leverage; following this logic, if clients tighten expansion budgets or financing conditions, new cloud providers’ order, project commissioning, and cash collection processes may be affected.

Thus, Rothschild’s key investment takeaway for the future performance and valuation of global AI new cloud compute rental players is that the market is likely to scrutinize valuations, financing capabilities, and capital returns for these two companies more closely. This is not a signal for order cancellations or downward operational guidance; in the months ahead, some investors may shift focus to both order growth and fund recovery abilities, and re-compare the return structures of compute operators, data center landlords, and power resource holders.

Stocks Focus on Growth, Credit on Commitments! Rothschild Says Credit Market Is “Issuing More Important Signals”

Rothschild & Co. Redburn states that new cloud companies’ profits and valuations are “cyclically inflated,” but remains “neutral” on the broader AI infrastructure sector from a cautious perspective.

Rothschild & Co. Redburn, closely associated with the Rothschild family in ownership and management, is the equity research and securities business unit of the Rothschild Group, serving institutional investors with research and execution services. The group is held by family members, senior partners, and other long-term investors, with Alexandre de Rothschild in overall charge.

On Monday, this institution took an even more cautious stance on global new cloud leaders Nebius (NBIS.US) and CoreWeave (CRWV.US), compared to bitcoin (BTC) miners transitioning into AI compute infrastructure. Both Nebius (NBIS.US) and CoreWeave (CRWV.US) received a “sell” rating due to the core argument that the credit market is pricing in risks largely ignored by the equity market for AI stocks.

Rothschild’s initial coverage of CoreWeave and Nebius assigns both a “sell” rating, while its first coverage of bitcoin miners transforming into AI compute infrastructure providers — Core Scientific (CORZ.US), TeraWulf (WULF.US), Applied Digital (APLD.US), Cipher Mining (CIFR.US), Hut 8 (HUT.US), and IREN (IREN.US) — is neutral.

Rothschild points out that the investment narrative in the equity market around generative AI infrastructure “remains firmly bullish,” but adds that “the credit market is beginning to price in risks largely ignored by the equity market.”

Against a generally strong market, NBIS shares rose as much as 1.1% in early US trading sessions, with bullish sentiment among retail investors on Stocktwits over the past day. CRWV shares rose about 1.8% and also maintained bullish sentiment among their community in the past day.

The credit market, however, is sending a markedly different and more cautious signal on AI. Rothschild believes the credit market is “issuing a more important signal,” since factoring in off-balance sheet commitments means hyperscale cloud providers have even higher actual economic leverage—a core logic that credit market caution demands investor attention.

The institution notes this could limit their ability to continue funding the next phase of large-scale AI infrastructure construction.

This warning is especially relevant to companies transitioning from bitcoin mining to AI and high-performance computing. CORZ, WULF, CIFR, HUT, and IREN are all developing data center businesses alongside, or in addition to, cryptocurrency mining, looking to monetize power resources and data center capacity with AI demand.

Among these companies, HUT shares led the gains with an intraday surge of 6%, followed by CIFR with a 4% jump. On Stocktwits, sentiment around HUT remained “bearish” over the past day, while sentiment for CIFR stayed “bullish.”

At the same time, IREN shares surged 3% intraday, WULF rose 2.5%, and CORZ about 2%. Over the past day, retail sentiment for IREN and WULF shifted from “neutral” to “bearish,” while sentiment for CORZ moved in the opposite direction.

Rothschild says that Nebius, backed by AI chip behemoth Nvidia (NVDA.US), faces the highest downside risk and that its share price already reflects the most optimistic outcome of continued revaluation of compute economics.

The institution believes the downside risk for CoreWeave is smaller than for Nebius, but its risk-reward profile “remains unattractive,” thus assigning a “sell” rating.

Companies directly involved in new cloud business models face greater valuation questions, while traditional real estate investment trusts may benefit indirectly from ongoing data center construction. Rothschild notes that traditional REITs, “as overlooked indirect beneficiaries, are especially worth watching.”

Bitcoin price rose in early trading, briefly approaching $85,000 before pulling back. Over the past 24 hours, the world’s largest cryptocurrency surged 5.5%, trading at about $84,800 at the time of writing. BTC remains the hottest ticker on the Stocktwits platform, with retail sentiment turning from “bearish” to “neutral” in the last day.

“New Cloud Forces” AI Compute Price Hikes Ignite Expectations, But Credit Markets Keep Pressing for Returns

Long-term compute procurement by hyperscale cloud providers not only support new cloud players’ orders but also create future payment obligations for customers. If end-user AI commercialization revenues fall short of prior expectations, yet rents, equipment upgrades, and financing expenses continue to be realized, pressure could be transmitted to CoreWeave, Nebius, and other new cloud companies via slower new orders, renegotiations on renewals, and counterparty credit risk. This is precisely why Redburn sees credit market signals as more important.

The seemingly endless expansion in AI inference-side computation demand still brings incremental business, but evaluation also depends on how much cash these increments ultimately generate. Thus, new cloud competition in the AI inference era will place greater importance on order quality, advance payment arrangements, hardware utilization, and capital returns after covering equipment renewal and funding costs.

The “credit market” as referenced by Redburn under Rothschild refers to the market where enterprises borrow money and investors price in debt repayment risks. This includes corporate bonds, bank loans, private credit, data center project finance, and risk hedging tools like credit default swaps (CDS). Stock investors focus on how much growth AI can deliver, while creditors are more interested in whether future cash flows can cover interest and principal. Watching the credit market mainly means looking at credit spreads—the extra yield investors demand for corporate bonds over similar-maturity government bonds—and the cost of CDS protection; if these indicators rise, it usually means investors are requiring higher risk premiums.

Redburn under Rothschild notes that equities are still aggressively pricing in AI growth, but credit capital is starting to account for repayment and funding pressure behind massive construction spending. According to published research summaries, the core assessment is: when off-balance sheet leasing and similar commitments are included, hyperscale cloud providers have higher actual economic leverage, potentially limiting their capacity to support the next wave of AI infrastructure expansion. Citing recent examples, Rothschild points to Oracle’s Project Jupiter data center, which secured an $18 billion loan priced at 89–91 cents on the dollar, reflecting investor caution over project execution and funding risks. Such severe credit constraints may be transmitted to CoreWeave and Nebius’ growth and valuation via customer expansion budgets, order realization, and financing conditions.

Redburn’s focus on the credit market echoes the latest warning from Michael Burry, famed for “The Big Short.”

According to reports, Burry estimates Amazon, Meta, Alphabet, Microsoft, and Oracle combined have nearly $1.2 trillion in uninitiated lease commitments, over $1.5 trillion in procurement obligations, and, considering special purpose entities, guarantees, and contingent support, total exposure of over $3 trillion. He makes clear he’s not alleging fraud; his main concern is that long-term contracts and procurement costs are locked in, but equipment values, technology requirements, and future demand could change faster, creating a mismatch for firms between capital recovery and hardware renewal.

It’s important to note that the $3 trillion mentioned by Burry is an aggregate estimate of various across-period commitments—it does not equate to realized losses; the investment implication is that, to assess the expansion capacity of AI data center giants, scrutiny of future cash outflows in financial footnotes is also necessary.