Is This the Big Short 2.0? What Happens to Global Markets If the AI Trade Unwinds

Michael Burry does not do things quietly. The man who bet against the entire US housing market in 2008 — the trade immortalised in The Big Short — has just disclosed a new basket of short positions. Nvidia. Applied Materials. Tesla. Caterpillar. The Philadelphia Semiconductor ETF. All of them, in a single Substack post published June 30, titled simply: Trading Post June 30, 2026.

“I have never shorted Caterpillar,” Burry wrote. “It has always done great for me on the long side in the past.”

That sentence should make you stop. Caterpillar is not a tech company. It builds bulldozers and mining equipment and power generators. But in 2026, investors decided it was an AI infrastructure play — because data centres need power, and power needs generators, and Caterpillar sells generators. The stock gained 86% in the first half of the year alone, reaching a price-to-sales ratio not seen in thirty years. Burry shorted it at $1,060.98. He thinks that number is absurd.

Whether Burry is right or wrong — and he has been spectacularly right once and spectacularly early many other times — his move forces a question markets have been avoiding for months: what actually happens if the AI trade unwinds?

Not in theory. In the real world — to indices, to commodities, to pension funds, to economies in South Korea and Taiwan that have quietly rebuilt their entire growth stories around this single thesis.

The Numbers That Worry People

Start with what Burry is actually looking at, because his argument is more specific than most bear cases. The Philadelphia Semiconductor Index — SOXX — is currently trading roughly 65% above its 200-day moving average. That extreme has only occurred once before in market history: at the peak of the dot-com bubble in 2000. The index’s price-to-sales ratio has exceeded 16 times. The S&P 500’s Shiller PE ratio sits above 40, its highest reading since the dot-com era.

Meanwhile the top ten AI stocks have gained 784% over the past twelve months. At the peak of the dot-com bubble, the equivalent cohort had gained 622%. By Burry’s math, the 2026 AI trade is running 162 percentage points hotter than even the most extreme phase of the previous tech bubble.

Then there is the capex question. Microsoft, Alphabet, Amazon and Meta have committed a combined $452 billion in AI infrastructure spending for 2026 alone. For a full breakdown of where AI stands today, see our State of AI Q3 2026 report. Nvidia has $119 billion in non-cancellable TSMC obligations. The bulls argue this spending is rational because AI is generating real, measurable revenue. The bears argue the issue is not whether AI works but whether the stocks have already priced in years of flawless execution that has not happened yet.

What a Correction Actually Looks Like

The phrase “AI bubble bursting” conjures images of 2000 — the Nasdaq falling 78% over two and a half years, dot-com companies vaporising, trillions of dollars of wealth erased. That framing is probably wrong for 2026, and it is worth being precise about why.

The 2000 crash was amplified by something specific: the companies at its centre had no revenue. They were burning cash on the promise of future monetisation that never arrived. The largest AI companies today are the opposite. Nvidia generated $81.6 billion in quarterly revenue in its most recent quarter, up 85% year-over-year. Microsoft, Google and Meta are profitable at a scale that had no equivalent in 2000.

This matters for the downside scenario. A correction in AI stocks is not the same as a collapse of AI companies. The more realistic bear case is a repricing — a period where the market decides that these earnings, while real and growing, do not justify multiples that assume everything goes right for the next decade.

The most realistic scenario is somewhere between the first two. A 25 to 35% Nasdaq correction from current levels would represent a painful but not catastrophic repricing. The S&P 500, less concentrated in AI names but still dominated by them, would likely fall 15 to 25%. These are meaningful corrections. They are not existential.

The Semiconductor Supply Chain: Where It Gets Complicated

Nvidia is the most visible target, but Burry’s short positions reveal something more interesting about where he thinks the real vulnerability lies. Applied Materials. The semiconductor ETF. Caterpillar. He is shorting the supply chain, not just the headline stock.

If hyperscalers cut AI capex — even modestly — the impact cascades backwards through the chip ecosystem faster than most people expect. Burry noted the math himself: if Microsoft cuts Nvidia capex by 20%, that equals a 4.2% Nvidia revenue hit. But Nvidia’s suppliers feel it first and worse. TSMC, which manufactures Nvidia’s chips and holds those $119 billion in non-cancellable commitments, would face margin pressure before Nvidia does. Applied Materials, which makes the equipment that makes the chips, operates further upstream and is historically more volatile than the chip companies themselves.

This is the semiconductor cycle playing out at an extreme scale. The industry is structurally cyclical — it always has been. Demand runs ahead of supply, companies over-order, inventory builds, orders get cancelled, stocks correct. What is different this time is the scale of the capex commitments and the speed at which they were made. The AI capex supercycle compressed what might have been a five-year build into eighteen months.

Copper, Gold, and What the Commodities Market Is Saying

Here is the angle most AI bubble pieces are missing: the commodity layer is already repricing, and that matters more than any individual stock move.

Copper was supposed to be one of the great beneficiaries of AI infrastructure. Data centres require enormous amounts of copper for wiring and cooling. The AI buildout was meant to drive a sustained copper supercycle — we covered the structural case for this in our piece on the global copper shortage. Instead, copper futures have quietly given back significant gains as the AI trade has shown volatility.

Gold and silver entered July in negative territory. Gold futures fell more than 1% in the first days of the month, while silver futures dropped nearly 3%. The reversal reflects rising rate expectations — central banks, particularly in the US under Fed Chair Kevin Warsh, are signalling a more hawkish stance than markets had priced.

Caterpillar’s 86% gain this year was built on exactly this assumption: that the physical infrastructure of AI would represent a decade of sustained demand. Burry’s short against it — his first ever — is a precise bet that this assumption has been priced too aggressively and too early.

The Global Contagion Map

An AI correction that stayed contained to US tech stocks would be painful but manageable. The more concerning scenario is one where the correction spreads through the global economy via specific channels.

South Korea and Taiwan are the most exposed. South Korea’s KOSPI is up 89% year-to-date, driven almost entirely by Samsung and SK Hynix on AI memory demand — a dynamic we explored in depth in South Korea’s KOSPI Is the World’s AI Hardware Barometer. Taiwan’s TSMC manufactures the most advanced AI chips in the world. A significant reduction in AI chip orders would hit both economies at a structural level — not just at the stock market level, but in exports, employment and government revenue. South Korea recently announced an 800 trillion won chip investment plan. Burry called that announcement “the beginning of the end.”

Japan is exposed through SoftBank, which has restructured itself as an AI investment vehicle, and through Arm Holdings, which it controls and which licenses chip architecture to virtually every AI semiconductor company.

European markets are exposed partly through Chinese consumer demand. A global pullback in AI investment sentiment typically flows through into Chinese consumer confidence and discretionary spending — which in turn affects brands like LVMH and Hermès that depend on Chinese buyers.

Energy markets are exposed because AI’s power demands have been a primary upward force on electricity prices and natural gas demand projections. If the buildout slows, energy infrastructure spending commitments face revision.

The Case Against the Bear

It is worth being honest about the counter-argument, because it is not trivial.

Burry has been early before — significantly early. He was bearish on the market in 2023 and it subsequently rose 131%. He has held bearish positions on Nvidia for over a year while the stock continued to climb. Being right about the structural argument and being right about the timing are two completely different things, and in a short position, the difference can cost you the trade.

There is also a scale argument. The hyperscalers are spending $452 billion because they are already generating returns from AI. This is not capital being allocated speculatively to technology that does not work. It is strategic spending by businesses with hundreds of billions in annual revenue on infrastructure that is already generating a return. A valuation correction is possible. A collapse of the underlying business is not.

This article represents the editorial view of AllinAllSpace and is based on publicly available information including SEC filings, Burry’s Substack posts, and market data as of July 3, 2026. It does not constitute investment advice. Sources include TheStreet, CNBC, Bloomberg, GuruFocus, IBTimes, and Investing.com. All figures cited are publicly reported.