Michael Burry warns Nvidia faces risks from distorted demand among limited buyers
The Big Short investor draws parallels to Cisco's dot-com collapse, arguing hyperscaler spending creates a dangerous 'bullwhip' effect in Nvidia's supply chain.
Share
Add us on Google by Editorial Team May. 26, 2026Michael Burry, the investor who became a household name for betting against the US housing market before the 2008 financial crisis, is now turning his skeptical eye toward the biggest winner of the AI boom. In a Substack post titled “The Heretic’s Guide to AI’s Stars Part III: Tracepalooza and the Bezzle,” Burry laid out a detailed case for why Nvidia’s current revenue trajectory may be built on shakier ground than most investors realize.
The concentration problem
Burry’s concern centers on Nvidia’s heavy dependence on hyperscale cloud providers. Microsoft, Google, Amazon, and Meta collectively account for roughly 50% of Nvidia’s data-center revenue.
To illustrate the fragility, Burry ran a simple stress test. A 20% reduction in Microsoft’s Nvidia-related capital expenditure alone could shave approximately 4.2% off Nvidia’s total revenue. In dollar terms, that translates to around $3.4 billion per quarter disappearing from the top line.
Nvidia’s latest quarterly revenue hit $81.6 billion, an 85% year-over-year increase. Burry drew a comparison to Cisco Systems during the dot-com era. Cisco was the picks-and-shovels play of the internet buildout, supplying the networking equipment that made the web possible. Its stock peaked in March 2000 and proceeded to lose nearly 90% of its value over the following two years.
AdvertisementThe bullwhip and the bezzle
Burry used the term “bullwhip effect” to describe what he sees happening in Nvidia’s supply chain: when a small change in end-consumer demand gets amplified as it moves up the supply chain, causing wild swings in orders and inventory at the manufacturer level.
Hyperscalers are currently in a frenzy of AI training and benchmarking, a phase that requires massive GPU purchases. But training workloads are inherently front-loaded. Once models are built, the ongoing compute needs for inference are different in both scale and hardware requirements.
The “bezzle” in his title refers to a concept from economist John Kenneth Galbraith: the period during a bubble when everyone thinks they’re richer than they actually are, before reality catches up.
Nvidia has committed to $119 billion in non-cancellable chip supply agreements, primarily with TSMC. Those commitments mean Nvidia is on the hook for massive volumes of silicon regardless of whether demand holds up.
Burry is putting money where his mouth is
Burry maintains significant short positions through put options on both Nvidia and Palantir. He has been raising concerns about AI-related spending and accounting practices since late 2025, making this the latest installment in a running thesis.
What this means for investors
When half your data-center revenue comes from four companies, your business model is essentially a bet on those four companies’ capital allocation decisions. Microsoft has already signaled some willingness to adjust spending plans based on returns, which makes Burry’s specific Microsoft stress test more than hypothetical.
The $119 billion in non-cancellable supply commitments adds another layer of risk. In a strong demand environment, those commitments are a competitive advantage. In a softening demand environment, they become an anchor. The difference between those two scenarios is entirely dependent on whether four or five companies keep writing checks at the current pace.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.
