Nvidia CEO Jensen Huang counters Michael Burry’s bear case with GPU rental data

Jensen Huang has a message for Michael Burry: the GPUs are renting just fine, thank you.

The Nvidia CEO pointed to H100 GPU rental rates surging 22% month-over-month to $3.28 per hour as of September 8, making a public case that his company’s chips are productive, durable assets rather than the speculative house of cards Burry has described.

The bear case: $105 billion in credit and a familiar-sounding alarm

Burry, who runs Scion Asset Management, has been raising red flags about Nvidia since at least late 2025. His core argument centers on what he calls “circular financing,” a claim that Nvidia is essentially helping fund the very customers who buy its chips.

The specific target of his criticism: Nvidia’s credit support tied to an OpenAI data center project in Ohio. According to Nvidia’s second-quarter 10-Q filing, that credit guarantee caps out at $105 billion, backing a facility with a 4.25 gigawatt IT load.

Burry has maintained short positions on Nvidia stock, betting that AI infrastructure spending is running ahead of actual returns, and that Nvidia sits at the center of that imbalance.

Huang’s rebuttal: the chips speak for themselves

Huang has framed the company’s GPUs as “fungible, durable and highly rentable,” treating them less like depreciating tech equipment and more like productive capital assets that generate ongoing revenue for their owners.

Nvidia CFO Colette Kress has argued that the company’s credit arrangements are designed to secure necessary infrastructure rather than artificially inflate chip sales.

On the depreciation question, Nvidia has pointed to actual GPU utilization patterns. The company says its chips typically remain in productive service for four to six years, which would make them longer-lived assets than critics assume.

Why this debate matters beyond Nvidia’s stock price

The $105 billion credit support figure is genuinely enormous. Even if the arrangement is structured as standard project finance, the concentration of risk is notable. If AI spending were to slow meaningfully, Nvidia would face exposure on both the revenue side, through reduced chip orders, and the credit side, through guarantees on projects that might struggle to generate sufficient returns.

For investors tracking the AI infrastructure cycle, the key metric to watch is exactly what Huang highlighted: GPU utilization and rental economics. If rental rates continue climbing, it validates the thesis that demand is real and sustainable. If rates plateau or decline, Burry’s warnings about overbuilt capacity start to carry more weight.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.