Someone will lose an enormous amount of money. Nobody knows who.
Michael Burry, the investor who foresaw the 2008 collapse, has turned his gaze toward artificial intelligence, warning that the accounting practices of major tech infrastructure companies are masking a bubble of historic proportions. His concern is not with the technology itself, but with the financial architecture surrounding it — the quiet manipulation of depreciation schedules that flatters profits while concealing risk. As in every era of speculative fever, the voices of caution and confidence speak simultaneously, and the market, for now, listens to neither completely.
- Burry accuses hyperscalers of artificially extending equipment depreciation timelines to inflate quarterly profits — a maneuver he compares directly to the accounting deceptions that preceded the 2008 financial crisis.
- 54% of fund managers surveyed believe AI stocks are already in a bubble, yet nearly 90% of those same believers are still actively investing in AI-related assets, revealing a market caught between fear and momentum.
- The Bank of England has raised its correction risk warning, while the IMF suggests any collapse would wound investors without destabilizing the broader financial system — a less catastrophic scenario than 2008, but painful nonetheless.
- Major figures remain genuinely divided: Bezos sees industrial bubbles as historically productive, Goldman Sachs calls AI infrastructure investment sustainable, and ABB's CEO warns not of a bubble but of a physical bottleneck in construction capacity.
- Sam Altman himself concedes that someone will lose an enormous amount of money — the uncertainty is not whether losses will come, but who will bear them when the enthusiasm finally meets its limits.
Michael Burry, the investor who predicted the 2008 housing collapse, is warning again — this time about artificial intelligence. After more than two years of silence on social media, he returned in late December to flag emerging bubbles, and has since named specific targets: Nvidia, Palantir, and the broader class of companies he calls hyperscalers, the cloud and data giants building out AI infrastructure at enormous scale.
His core accusation is accounting-based. When companies buy hardware with a two-to-three-year product cycle, standard practice would depreciate it over that same window. Burry argues hyperscalers are stretching those timelines artificially, reporting lower expenses and inflating profits in a way he calls one of the most common frauds of the modern era — an echo of the creative accounting that preceded 2008.
The market's response is divided. A Bank of America survey found 54% of fund managers believe AI stocks are already in a bubble, yet UBS noted that roughly 90% of those same skeptics are still investing in AI anyway. The Bank of England has warned of growing correction risk, while the IMF's chief economist suggests any collapse would resemble the dot-com bust but remain less systemically dangerous, since this cycle is far less debt-financed.
Among major voices, Jeff Bezos acknowledged that high enthusiasm funds bad ideas alongside good ones, but argued industrial bubbles often leave useful inventions behind when they pop. Goldman Sachs called AI infrastructure investment sustainable, while ABB's CEO saw not a bubble but a bottleneck — the physical capacity to build all this infrastructure simply doesn't exist yet, and won't for years.
Sam Altman was candid: yes, investors are too enthusiastic, and someone will lose an enormous amount of money. Someone else will make an enormous amount. Nobody knows which side they're on. Burry's warnings carry the weight of a proven track record, but whether they signal prophecy or pessimism is the question the market has not yet answered.
Michael Burry, the investor who saw the 2008 housing collapse coming before almost anyone else, is warning again. This time his target is artificial intelligence—specifically, the massive sums pouring into AI infrastructure and the accounting tricks he believes are masking a bubble about to burst. In late December, after more than two years of silence on social media, Burry posted on X that he saw bubbles forming. By this week, he was naming names: Nvidia, Palantir, and the entire class of companies he calls "hyperscalers" that buy and deploy AI chips and servers.
His specific complaint cuts to the heart of how these companies report their finances. When a company buys expensive hardware with a product cycle of two to three years, standard accounting would depreciate that equipment over that same timeframe. But Burry argues that hyperscalers—the massive cloud and data companies building out AI infrastructure—are artificially extending the useful life of their equipment on the books. The effect is simple: longer depreciation schedules mean lower expenses reported each quarter, which inflates profits. He calls it one of the most common frauds of the modern era, a sleight of hand that echoes the creative accounting that preceded 2008.
The market itself seems divided on whether he's right. A monthly survey by Bank of America Global Research found that 54 percent of fund managers believe AI stocks are already in a bubble. The other 46 percent disagree. The Bank of England, meanwhile, has sounded an alarm: the risk of a sharp market correction has grown. Yet despite these warnings, investors keep buying. UBS equity strategists noted in October that roughly 90 percent of the people who say they believe we're in an AI bubble are still actively investing in AI-related areas anyway.
The voices of major economic figures reveal the genuine uncertainty. Jeff Bezos, speaking in Italy in early October, acknowledged that when enthusiasm runs this high, bad ideas get funded alongside good ones. But he drew a distinction: industrial bubbles, he argued, aren't necessarily destructive. When they pop, society keeps the useful inventions. Bryan Yeo, who oversees Singapore's sovereign wealth fund, identified a smaller speculative bubble in AI startups—companies slapped with an AI label and valued at enormous multiples of their actual revenue. Joseph Briggs at Goldman Sachs countered that the flood of investment into U.S. AI infrastructure is sustainable, though he cautioned that the ultimate winners remain unclear as technology shifts rapidly and switching costs stay low.
Morten Wierod, CEO of ABB, offered a different concern altogether. He doesn't see a bubble, but he sees a bottleneck: the physical capacity to build all this infrastructure doesn't exist yet. We're talking about trillions in planned investment, he said, and it will take years to construct because there simply aren't enough people and resources to do it. The International Monetary Fund's chief economist, Pierre-Olivier Gourinchas, suggested that an AI investment collapse could resemble the dot-com bust but would likely prove less systemically dangerous. The dot-com crash was financed heavily with debt; this one isn't. If the market corrects, shareholders and investors lose money, but the broader financial system stays intact.
Sam Altman, CEO of OpenAI, was perhaps the most candid. Yes, he told The Verge in August, investors are too enthusiastic about AI right now. Someone will lose an enormous amount of money. Someone else will make an enormous amount. Nobody knows who. President Trump, asked directly about a bubble, said no—he loves AI and thinks it will be very useful. The market, for now, keeps climbing. Burry's warnings echo from the past, but whether they'll prove prophetic or merely pessimistic remains the question that keeps investors awake at night.
Notable Quotes
Extending the useful life of assets artificially to boost reported earnings is one of the most common frauds of the modern era.— Michael Burry, investor
Industrial bubbles aren't necessarily destructive—when they calm down, society keeps the useful inventions.— Jeff Bezos, Amazon founder
Yes, investors are too enthusiastic about AI right now. Someone will lose an enormous amount of money.— Sam Altman, OpenAI CEO
The Hearth Conversation Another angle on the story
Why does Burry focus on depreciation schedules? That seems like an accounting detail, not the core of a bubble.
Because that's where the fraud lives. If you extend how long equipment stays on the books as an asset, you report lower depreciation expenses. Lower expenses mean higher profits. Higher reported profits justify higher stock prices. It's not creating value; it's hiding the true cost of doing business.
But couldn't companies legitimately extend equipment life if the technology actually lasts longer than expected?
They could. But Burry's point is that they're doing it systematically, across the board, to smooth earnings. It's not based on actual equipment longevity—it's based on what the spreadsheet needs to look good.
So why do 90 percent of investors who think it's a bubble keep investing anyway?
Because they don't know when it pops. Being right about a bubble and being right about timing are two different things. You can believe the market is overheated and still lose money betting against it too early. Easier to ride the wave and hope you're out before it crashes.
What's the difference between this and the dot-com bubble?
Debt. The dot-com crash was financed with borrowed money, which meant when it collapsed, it threatened the whole financial system. This AI spending is mostly equity-funded. When it corrects, investors lose their money, but banks don't fail. The damage is real but contained.
Is there any scenario where Burry is wrong and this isn't a bubble at all?
Yes. If the productivity gains from AI actually materialize at the scale being promised, if the infrastructure spending pays off in real economic growth, then the valuations might be justified. But that's a big if, and it requires the technology to deliver on promises that haven't been proven yet.