Current market focus: Is AI capital expenditure "peaking" or undergoing a "mid-term correction"?
Jefferies believes that the market has priced in a "peak" scenario, but there is no "real-world evidence" to support this judgment: Nvidia, with an off-balance-sheet exposure of $330 billion, acts as the "central bank" of the AI ecosystem, and a shift in governance could tighten the financing chain; the net profits of the four major cloud providers grew by 73%, but only by 15% after excluding non-cash AI-related gains; new chip export regulations in Southeast Asia have further raised the threshold for cross-border expansion of computing power.
The behavioral patterns of the financial markets are increasingly resembling those seen when the AI capital expenditure cycle peaks, even though there is currently no convincing evidence on the ground to support this view. According to Jefferies, the core divide in today’s market is whether this is truly the peak of the AI capital expenditure cycle or just a mid-cycle correction during an ongoing uptrend.
According to Quwind Trading Desk, Jefferies stated in its latest research report released on November 19 that multiple risk signals are simultaneously fermenting around this core assessment.
Nvidia’s “central bank” role in the AI ecosystem is coming under governance pressure; should its enormous off-balance-sheet financial exposures contract, this could exert a systemic tightening effect across the entire AI financing chain;
Meanwhile, the quality of hyperscaler cloud providers’ earnings is being scrutinized more strictly by the market—after excluding non-cash mark-to-market gains from AI-related investments, their core profit growth is far below the headline numbers;
Additionally, the US Department of Commerce's Bureau of Industry and Security (BIS) announced new licensing requirements for AI chip exports to several Southeast Asian countries and Brazil, further reshaping the geographic distribution and expansion pace of global compute capital expenditures.
The bank believes these three lines of logic are intertwined, constituting the core controversy in qualifying the current AI capital expenditure cycle.
Market Signals and Cycle Definition: Pricing Leads “Ground-Level Evidence”
Jefferies' report makes it clear that the financial markets’ behavior now increasingly resembles the peak of the AI capital expenditure cycle, but there is still a lack of convincing “ground evidence” for this judgment.
Continuing the analytical approach from its November 5 report, Jefferies temporarily gives priority to market signals, framing the current situation as a contest between two scenarios:
First, this is the true peak of the AI capital expenditure cycle; second, this is merely a mid-cycle correction within an ongoing uptrend.
The report considers this judgment to be unsettled in the absence of additional fundamental evidence and identifies it as the most crucial qualitative issue currently facing investors.
Nvidia’s “Central Bank” Role: Massive Off-Balance-Sheet Exposures & Governance Shift Risks
Nvidia has played the role of “financier” and “guarantor” of the entire AI ecosystem over the past two years, with the scale of its financial exposures far exceeding the market’s previous general understanding. According to a summary posted by Tomasz Tunguz on November 12, Nvidia’s commitments include:
$42 billion for technology acquisition commitments;
$61 billion in equity investments, spanning 229 startups and 13 public companies;
$164.5 billion in lease guarantees and cloud compute capacity backstops, requiring corresponding capital reserves;
$63.1 billion in accounts receivable from customers, with payment terms extending up to a year.
Changes in governance are drawing market attention. Colette Kress, who has served as Nvidia’s CFO for 13 years and is now the interim CEO, has expressed reservations about these massive financial risk exposures. The report argues that if Nvidia shifts from an aggressive matchmaker model to more conservative financial discipline, as the leading player in AI deals, it will clearly exert a tightening and cascading effect on the entire AI capital expenditure and financing chain.
This pressure is particularly pronounced for “neoclouds.” Nvidia previously viewed “neoclouds”—i.e., data center startups—as a key alternative source of compute demand outside the hyperscale cloud providers. Worth noting is that all major hyperscalers, such as Alphabet with its TPU, are advancing proprietary chip efforts.
As the related turmoil evolves, CoreWeave’s stock price has pulled back 25% since the news broke, with its USD bond yield rising to around 15.7%, reflecting significant credit and survival pressure.


The report concludes that if the most speculative segment of compute demand contracts, it is likely to drive down compute prices for the remaining participants, creating repricing pressure across the entire AI compute supply and demand structure.
Hyperscaler Profit Decomposition: Reflexivity Warnings for Capital Return on Investment
Jefferies’ report conducted a deep breakdown of the profit structure of Microsoft, Amazon, Alphabet, and Meta, exposing how earlier profit growth for hyperscalers has been heavily reliant on non-cash mark-to-market gains from AI-related investments, rather than substantial improvements in core operating profits.
The “other non-operating income and loss” tracked by the report mainly includes unrealized gains and losses from unlisted AI-related investments and other non-cash items. This quarterly total swung sharply from a positive $155 billion in 2Q26 to a negative $9 billion in 3Q26, directly reflecting mark-to-market reversals as AI asset prices fell.

On an annualized basis, the data is equally noteworthy: In the 12 months ending 3Q26, the four tech giants' combined annualized net profit rose 73% from the preceding 12 months ending 1Q25, a $248 billion increase (from $340 billion to $588 billion). However, of this increase, annualized “other non-operating income” soared 18-fold from $12 billion to $209 billion, contributing the vast majority ($198 billion) of the profit growth; after deducting this item, the remainder of annualized net profit only grew 15%, or $51 billion (from $329 billion to $379 billion).

The report notes this highlights how the prior AI bull market’s “reflexivity” amplified reported profits and is making markets even more sensitive to the real returns of massive capital expenditure. As AI asset prices adjust downward, this “reflexivity” is operating in reverse.
At the same time, the US Department of Commerce's Bureau of Industry and Security (BIS) has announced export license requirements for AI chips to Southeast Asian countries and others.
The triple pressures of tighter export controls, heightened regulatory scrutiny, and a tighter funding environment mean that cross-border compute expansion and data center projects relying on external capital will face higher compliance and financing barriers. The report believes this will affect the geographic distribution and slope of global AI capital expenditure expansion.
Disclaimer: The content of this article solely reflects the author's opinion and does not represent the platform in any capacity. This article is not intended to serve as a reference for making investment decisions.
You may also like
Anthropic adds another victory in listing, Nasdaq (NDAQ.US) further distances itself from the New York Stock Exchange
Anthropic's listing further consolidates Nasdaq's leading position over its rival, the New York Stock Exchange.
On Anthropic's first day of listing, retail trading volume hit a record high, but nearly half of the trades were sells!
On its first day of listing, Anthropic saw a total retail trading volume of $772 million, setting a new historical high. However, nearly half of the transactions were sell orders, with net buying amounting to only $60 million, far below the expected $250 million. Dragged down by the overall market decline, a high proportion of retail profit-taking, and funds diverted in advance to crypto perpetual contracts, the stock opened high but closed lower. Going forward, it still faces potential selling pressure from a significant amount of low-priced retail holdings.
Shorts are running wild, gold urgently needs a catalyst to break the tightening narrative!
Deutsche Bank has issued a warning that macro funds and CTA strategies have maintained "maximum short positions" in gold for several weeks, and there is still ample room for further shorting. The key contradiction lies in the fact that ETF holdings remain stable, central banks continue to increase their gold reserves, and there are no signs of physical selling. However, financial positioning effects have completely disconnected the price from fundamentals. Analysts believe that gold’s sensitivity to the US dollar is approaching levels not seen since 1972, and technical analysis points to $3,500. A resolution will require one of three catalysts: a dovish turn by the Federal Reserve, weakening economic data, or cracks appearing in the broader market.
Ross Stores (ROST.US) Q3 Conference Call: 8% Same-Store Sales Growth on a High Base, CEO Says Traffic Drives an Extremely Healthy "Growth Flywheel"
After releasing its third-quarter fiscal 2026 results (ending October 31) after the U.S. market closed on November 19, Ross Stores, the American discount apparel and home goods retailer, held an earnings conference call.
