Shares across the AI infrastructure complex are trading mixed after well-known short seller Jim Chanos publicly compared today’s artificial intelligence capital spending boom to the 1999-2000 telecom build-out. NVIDIA (NASDAQ:NVDA | NVDA Price Prediction) stock is on the radar, along with several names Chanos flags as “equipment leasing.” Among the potentially vulnerable stocks in this discussion are CoreWeave (NASDAQ:CRWV), Nebius Group (NASDAQ:NBIS), Taiwan Semiconductor Manufacturing (NYSE:TSM), and Vistra (NYSE:VST).
The man issuing the warning has a notable track record. Chanos founded Kynikos Associates and built his reputation on calling out Enron before it collapsed. His “same setup” framing has investors revisiting valuations across AI compute, cloud, foundry, and power. Broadcom (NASDAQ:AVGO) sits in the same vendor camp the thesis touches.
Chanos Argues It’s the Same Dot-Com Setup
In an iConnections interview summarized by Podcast Alpha, Chanos walks through how S&P 500 earnings rose 30% over two years from 1998 to 2000, then fell 40% in 12 months during a mild recession. His point is that the earnings swing was not really driven by the recession itself. The culprit was the telecom buildout collapsing as companies realized they had ordered roughly 10,000 routers when they needed 2,000.
According to Chanos, the AI capital expenditure boom has “identical mechanics.” He contends that NVIDIA’s earnings are “CapEx boom output” and that S&P 500 estimates are rising fast because infrastructure spending flows directly into a small number of vendors. If order books get pulled, the snapback could arrive quickly.
Since that iConnections appearance, Chanos has sharpened his warning. Speaking on the Prof G Markets podcast in late July 2026, he said the current moment feels “closer to a ’99-type moment than a ’97,” placing the AI boom in the final stretch before the dot-com collapse rather than the middle of a still-healthy cycle. He also flagged a surge in new stock issuance, calling 2026’s IPO pipeline “an entirely different animal” compared to earlier in the boom, and pointed to an accounting mismatch: companies capitalizing AI spending and depreciating it over five to ten years, while suppliers like NVIDIA recognize the same spending as revenue immediately.
The bullish counterargument carries real weight. NVIDIA just reported Q1 fiscal 2027 revenue of $81.6 billion, up 85% year over year, with Data Center revenue of $75.2 billion growing 92% in the same period. CEO Jensen Huang stated that “the buildout of AI factories, the largest infrastructure expansion in human history, is accelerating at extraordinary speed.” NVIDIA also carries $119 billion in supply-related commitments and guided Q2 FY2027 revenue to approximately $91 billion, suggesting demand visibility that extends well beyond a single quarter.
His Specific Targets: CoreWeave, Nebius, and Power Plays
Chanos reserves his sharpest critique for AI cloud lessors. He argues that CoreWeave and Nebius are essentially equipment-leasing businesses generating mid-to-high single-digit pre-tax returns on capital, and that neither should trade at higher multiples than NVIDIA and Taiwan Semiconductor, which actually control GPU supply.
The numbers are striking. CoreWeave posted Q1 2026 revenue of $2.08 billion, up 112% year over year, yet also reported a $740 million GAAP net loss for the quarter, a return on equity of negative 41%, and capital expenditures of $7.7 billion in a single quarter. Its revenue backlog surged from $25.4 billion to $99.4 billion over the past year, with the company signing more than $40 billion of new commitments in Q1 alone. Nebius trades at a price-to-earnings ratio of roughly 100x against negative EBITDA. Chanos argues these valuations are structurally misaligned given the companies’ role as intermediaries rather than supply controllers.
On the power side, Chanos contends that alternative-energy stocks trading at 50 to 70 times earnings as data-center power plays will revert once U.S. permitting bottlenecks resolve within two to three years. His logic is that power represents only 5% to 7% of data center revenues, so the premium applied to power names reflects hype rather than structural pricing power. Vistra trades at a forward P/E of roughly 26x and has signed long-term power purchase agreements with Meta Platforms (NASDAQ:META) at PJM Interconnection nuclear sites, which provides some contractual insulation from the thesis.
SpaceX Valuation and the Broader Critique
Chanos extends the bubble framing beyond public AI names. He cited SpaceX’s IPO as emblematic of stretched expectations, pointing to a roughly $2 trillion valuation against 2025 revenue of approximately $18.6 billion. On that basis, investors are pricing in business models, including orbital AI infrastructure and next-generation launch services, that do not yet generate meaningful revenue. That gap between present cash flows and implied valuation is, in Chanos’s view, the defining feature of a late-stage bull market in AI.
SpaceX (NASDAQ:SPCX) has genuine assets. Starlink generated approximately $11.4 billion in revenue in 2025, accounting for roughly 61% of the company’s total, and surpassed 10.3 million active customers across 160 countries as of March 2026. But Chanos’s argument is not that the core business lacks value. It is that the premium layered on top of Starlink, attributable to Starship, xAI integration, and other speculative bets, represents the same “build it and they will come” logic that defined the fiber-optic overbuilding of 1999.
It is worth emphasizing that this is one prominent skeptic’s thesis. Timing short calls is notoriously difficult, and AI demand has been durable so far. Chanos himself acknowledged on the Prof G Markets podcast that the cycle may have more runway before the inflection arrives.
What to Watch From Here
The next concrete catalyst is NVIDIA’s Q2 FY2027 earnings call, now confirmed for August 26, 2026. Any sign of softening order books or reduced hyperscaler commitments in that report would give Chanos’s thesis immediate market relevance. Broadcom has guided Q3 FY2026 AI semiconductor revenue to $16 billion, representing over 200% year-over-year growth, and also projected full-year fiscal 2026 AI chip revenue of $56 billion, up roughly 180% from fiscal 2025. Those numbers will test the demand narrative directly when Broadcom reports its own Q3 results in September.
For now, the market is not endorsing the Chanos call. Watching whether AI cloud lessors keep raising capacity commitments, and whether NVIDIA’s supply book translates into the customer concentration that worries short sellers, will be the cleanest signal of whether this cycle is still accelerating or beginning to turn.
Editor’s note: This update corrects NVIDIA’s next earnings call date to August 26, 2026 (from August 27), refreshes NVIDIA’s Q1 FY2027 figures to $81.6 billion in total revenue and $75.2 billion in Data Center revenue (up 92% year over year), adds CoreWeave’s Q1 2026 revenue of $2.08 billion and its backlog growth from $25.4 billion to $99.4 billion, incorporates Broadcom’s full-year fiscal 2026 AI semiconductor revenue target of $56 billion, and includes Chanos’s more recent July 2026 Prof G Markets warning that the cycle is now closer to a “1999-type moment.”
Contact [email protected] for any questions or corrections.
Top Gaining Stocks
Top Losing Stocks