1 hour ago
21
Michael Burry, the investor famous for betting against the US housing market before the 2008 financial crisis, is now sounding alarms about Nvidia. His argument: the chipmaker’s extraordinary revenue growth is being powered by a small group of buyers whose spending habits may not last.
In a Substack post published on May 24, titled “The Heretic’s Guide to AI’s Stars Part III: Tracepalooza and the Bezzle,” Burry laid out what he sees as the structural fragility underneath Nvidia’s headline numbers. The core thesis is that demand from hyperscalers like Microsoft, Google, Amazon, and Meta is temporarily distorted, not permanently elevated.
The numbers look great, until you squint
Nvidia’s most recent quarterly revenue came in at $81.6 billion, representing 85% year-over-year growth.
Hyperscalers account for approximately 50% of Nvidia’s data-center revenue, according to Burry’s analysis. That level of concentration means a relatively small shift in purchasing behavior from just one or two customers could ripple through the entire business.
Burry ran a specific scenario to illustrate the point. He estimates that a 20% reduction in Microsoft’s capital expenditures on Nvidia chips alone would translate to roughly a 4.2% decrease in Nvidia’s total revenue. Applied to an $81.6 billion quarter, you’re talking about billions of dollars evaporating from a single customer pulling back by a fifth.
The bullwhip effect and $119 billion in commitments
Nvidia currently holds $119 billion in non-cancellable supply commitments, primarily with TSMC, the Taiwanese foundry that manufactures its chips. Nvidia has locked itself into buying an enormous amount of semiconductor production capacity that it cannot walk away from, regardless of whether customer demand holds up.
Burry frames this through the lens of the “bullwhip effect,” a well-documented supply chain phenomenon where small fluctuations in end-customer demand get amplified as they move upstream through the supply chain. A 10% dip in actual AI workload demand could trigger a 20% or 30% swing in chip orders as each link in the chain overreacts.
For Nvidia, $119 billion in non-cancellable commitments is a very large bet that the current pace of AI infrastructure spending continues without a meaningful pullback.
What the hyperscalers are actually doing
Burry’s post characterizes the current hyperscaler spending phase as “transitory.” He argues that Microsoft, Google, Amazon, and Meta are in a period of excessive capital deployment focused on training AI models, benchmarking performance, and harvesting data at scale.
Burry has maintained a bearish outlook on AI infrastructure since late 2025 and continues to hold put options on Nvidia and related stocks.
What this means for investors
The customer concentration risk is real and quantifiable. When half your data-center revenue comes from a handful of companies, you’re not really a diversified business.
The $119 billion in non-cancellable TSMC commitments adds a layer of rigidity that could become painful in a downturn. If hyperscaler spending decelerates even modestly, Nvidia could find itself with committed supply that exceeds actual demand.
For investors holding Nvidia or broadly exposed to AI infrastructure plays, the key variable to watch is hyperscaler capital expenditure guidance. Microsoft, Google, Amazon, and Meta all report CapEx figures quarterly, making this one of the more trackable macro risks in the current market.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.
English (US) ·