Burry: Nvidia's Real Risk Is $3 Trillion in AI Spending, Not Huang

Michael Burry estimates Big Tech has committed more than $3 trillion to future AI spending, and he argues that bill, not the leadership crisis at Nvidia, is the real threat to the company. His argument landed the same day the ground shifted under the chipmaker: on November 12, 2026, Jensen Huang resigned as CEO, President, and member of the Board of Directors, effective immediately. CFO Colette Kress took over as interim CEO while keeping her CFO role.

The stock is absorbing both shocks at once. Nvidia traded at $166.92 late that morning, down 3.92% on the day, 16.59% over the past week, and 26.25% over the past month. Burry, the Scion Asset Management founder made famous by The Big Short, says investors are watching the wrong problem.

What does Burry actually claim?

In a Substack post reported by Business Insider, Burry said he was "a bit shocked at how the market has turned on the company." He raised questions about whether the "scandal will hurt the economics of the business tremendously." That phrasing is a question, not a verdict, and Burry did not quantify the damage.

His bear case is structural, not personal. He warned that Nvidia's competitive advantage could shrink faster over the next five years than he had previously expected. He also described Nvidia's investments in other AI players, and the demands it attaches to them, as "mafia-like behavior." That language is strong; Burry offered no named examples in the reported excerpts.

The core of his argument is depreciation accounting. Hyperscalers (the largest cloud providers such as Amazon, Microsoft, and Google) spread the cost of GPUs over many years through depreciation, the accounting charge for wear and obsolescence. Burry argues aging chips could lose value faster than customers can earn back what they paid for them. If the depreciation schedule runs longer than the chips' real working life, reported profits today are too high, and future purchases could slow once the true costs surface.

The takeaway in plain words: Burry thinks the AI build-out is being financed by optimistic accounting, and the correction will show up in purchase orders, not press releases.

What do Nvidia's own filings show?

Nvidia's most recent results were strong. The Q2 fiscal 2027 release, filed with the SEC, reported revenue of $96.221 billion, up 105.85% year over year. Data Center revenue was $89.023 billion. Non-GAAP EPS of $2.22 beat the $2.0887 consensus.

The same filing documents how much risk Nvidia has absorbed on behalf of its customers:

• Supply obligations of $279.00 billion, mostly memory purchases for the Vera Rubin platform.
• Guarantees to AI cloud and data center partners capped at $108.5 billion.
• Nearly $50 billion invested in frontier AI labs.
• Days sales outstanding (the average time to collect payment) rising to 60 days from 45 days.

Those numbers give Burry's circularity critique its substance. Some financing deals include a "take-or-pay commitment on a portion of the facility's capacity, a minimum revenue guarantee." Nvidia also invested $2 billion in CoreWeave (NASDAQ:CRWV), one of the AI cloud companies that buys its chips.

Kress addressed the criticism directly on the earnings call: "We know some will call this circular financing. We see it differently." She did not, in the reported remarks, dispute the underlying figures.

The takeaway: Nvidia is simultaneously the supplier, the financier, and the shareholder in parts of its own demand chain, and the filing quantifies that exposure at hundreds of billions of dollars.

How does management answer the obsolescence question?

Management's response to the chip-lifespan concern is that its hardware is "fungible and durable and can be redeployed to support other customers." Huang, before his resignation, cited a payback period of less than a year for some data centers. He also noted that demand runs far above the roughly 70% fiscal 2028 growth that supply can currently support.

Nvidia expects capital spending by the top five hyperscalers to reach nearly $800 billion in 2026 and $1.3 trillion in 2027. That forecast rests on customers continuing to spend. AI labs are expected to make up roughly a quarter of Nvidia's business next year, and Kress said these labs are growing faster than their balance sheets and credit profiles can support. At a P/E of about 35, investors are paying for that spending to continue uninterrupted.

The caveat writes itself: management's payback math assumes redeployment demand exists at scale, which is precisely what a demand slowdown would test first.

The takeaway: the bull case depends on customers Nvidia itself is helping finance, and management has acknowledged the strain on those customers' credit.

What should investors watch at Tuesday's earnings?

Nvidia will report earnings on Tuesday as scheduled. The guidance on record calls for revenue of $108.0 billion, plus or minus 2%, and a gross margin of 74.0%.

Three metrics will test Burry's thesis against management's:

• Whether the guarantees and supply obligations keep growing.
• Whether receivables are collected faster or slower, measured in days sales outstanding.
• Whether gross margin reaches its expected Q4 low of 71% to 72% on schedule.

Burry also flagged a risk outside Nvidia's control: a potential US government crackdown on the company's chip sales to China. That adds a policy dimension to what is otherwise a debate about accounting and demand durability.

Yet Burry is not uniformly negative on the trade. He said he could "imagine the whole market is going to think this is a heck of a trade fairly soon." The same post that warns of depreciation risk and "mafia-like" deal terms also leaves room for the stock to rally.

The takeaway: Tuesday's numbers will not settle the debate, but the trajectory of receivables, guarantees, and margins will show whose model of the AI boom is holding.

The stakes extend past one stock. If Burry is right about depreciation schedules outrunning chip lifespans, the correction would hit the capital spending plans of every hyperscaler counting on those GPUs to pay for themselves, and the $1.3 trillion spending forecast for 2027 would become the number to distrust.