Key Takeaways:
Short seller Jim Chanos says the AI infrastructure buildout has "identical mechanics" to the 1999-2000 telecom collapse that erased $5 trillion in market value.
Jim Chanos, the short seller who called Enron, compared the artificial intelligence capital expenditure boom to the 1999-2000 telecom buildout, sending AI infrastructure stocks into mixed trading Wednesday.
"We have the same setup as the dot-com crash," Chanos, founder of Kynikos Associates, said in an iConnections interview. "Nvidia's earnings are CapEx boom output."
Chanos argued that S&P 500 earnings rose 30% over two years from 1998 to 2000, then fell 40% in 12 months during a mild recession — not because of the recession, but because telecom companies realized they had ordered roughly 10,000 routers when they needed 2,000. He contends the AI infrastructure cycle has "identical mechanics," with spending concentrated among a small number of vendors.
If order books get pulled, the snapback could be quick. Nvidia reports earnings Aug. 27, and Broadcom's Q3 AI semiconductor revenue guidance of $16 billion — representing over 200% year-over-year growth — will test whether demand can sustain current valuations.
Chanos targets AI cloud lessors and power plays
The short seller reserved his sharpest critique for AI cloud lessors, arguing that CoreWeave and Nebius Group are essentially equipment-leasing businesses generating mid-to-high single-digit pre-tax returns on capital. CoreWeave trades at a price-to-sales ratio of 9.35 with a return on equity of negative 41% and a $740 million net loss in the first quarter of 2026, according to its filings. Nebius trades at a price-to-earnings ratio of 100 times against negative EBITDA. Chanos said neither should trade at premium multiples to Nvidia and Taiwan Semiconductor Manufacturing, which control the GPU supply.
On power, Chanos argued that alternative-energy stocks trading at 50 to 70 times earnings as data-center power plays will revert once U.S. permitting bottlenecks resolve in two to three years. Power represents only 5% to 7% of data center revenues, he noted. Vistra, which trades at 26 times earnings, has signed long-term power purchase agreements with Meta Platforms at PJM Interconnection nuclear sites.
The bull case and what to watch
The bullish counterargument is substantial. Nvidia posted 85% year-over-year revenue growth with data center revenue of $75.25 billion, and Chief Executive Officer Jensen Huang has called the buildout "the largest infrastructure expansion in human history." Nvidia also holds $119 billion in supply-related commitments, suggesting visibility well beyond a single quarter.
CoreWeave's backlog has swelled to $99.4 billion, and Broadcom's AI semiconductor revenue is on track to more than double this fiscal year. The bull case rests on bookings holding; the bear case rests on them not.
Chanos extended his bubble framing beyond public AI names, noting that SpaceX went public at roughly 90 times revenue on a roughly $2 trillion valuation, with core Starlink mobile supporting at most a couple hundred billion in value. That leaves roughly $1.5 trillion tied to business models that do not yet exist, he argued.
For investors, the key catalysts are Nvidia's Aug. 27 earnings call, where any hint of softening order books would pressure the entire AI complex, and Broadcom's Q3 results, which will test whether triple-digit growth rates can persist. Nvidia shares trade at about 35 times forward earnings, a premium that leaves little room for execution missteps.
This article is for informational purposes only and does not constitute investment advice.
