Michael Burry, the investor made famous by The Big Short, is once again stirring up Wall Street. This time, he’s taking aim at the tech giants fueling the AI boom companies like Meta, Oracle, Microsoft, Google, and Amazon and warning that their financials might be far less solid than they look. His latest criticism centers on how these so-called hyperscalers handle one of the most fundamental accounting measures in business: depreciation. According to Burry, the world’s biggest tech firms are inflating their earnings by stretching out how long their expensive AI hardware is supposed to last.
In a post on X, Burry argued that hyperscalers are aggressively under-reporting depreciation expenses on billions of dollars of data-center hardware, GPUs, and servers that make up the backbone of AI computing. The argument is simple but dangerous: if you assume these machines last six to eight years when they really burn out in two or three, your profits look much higher than they actually are. Burry estimates that between 2026 and 2028, companies like Meta and Oracle could collectively understate depreciation by nearly $176 billion. That would mean Oracle’s earnings could be inflated by around 26 percent and Meta’s by about 20 percent.
The implication is massive. These numbers suggest that much of the profit growth being celebrated in AI-related stocks may be more illusion than reality. It’s not that the tech giants are cooking the books outright it’s that the accounting assumptions they use make their operations appear far more efficient and sustainable than they might be. For Burry, that smells like a bubble: shiny on the surface, hollow underneath.
What makes this critique even more striking is that it echoes a warning issued decades ago by another legendary short seller: Jim Chanos, the man who called Enron’s collapse. Chanos argued that when companies disguise the true cost of their assets or obscure how quickly their investments lose value, investors end up believing in a growth story that doesn’t exist. Enron’s downfall was fueled by complicated accounting tricks that hid the decay of its core business behind numbers that looked consistently good. Burry’s point is that hyperscalers might not be committing fraud, but they’re using accounting optimism to build an empire that could start cracking once the replacement costs of AI infrastructure hit full force.
AI hardware is not like a power plant or an office building it ages fast. The pace of chip innovation is brutal. Today’s cutting-edge GPU can be obsolete within two years, replaced by something cheaper, faster, and more efficient. If companies are depreciating those assets over long timelines, their balance sheets are painting an unrealistic picture. Burry warned that when these firms are forced to refresh their hardware sooner than their accounting assumes, the impact on earnings could be sudden and severe. Investors who bought into the AI-driven profitability story might find themselves facing a harsh reset.
Jim Chanos had said something similar earlier this year, pointing out that hyperscalers were spending hundreds of billions on new data centers without a clear line of sight to monetization. He warned that this could mirror past bubbles, where investors ignored the math because the growth narrative sounded too good. Burry’s new analysis essentially puts numbers to that same thesis. Both men see the AI boom as a potential trap for investors who assume the big tech players can spend endlessly without paying the price later.
Supporters of these companies push back hard. They argue that AI infrastructure is a long-term investment, that hardware can be upgraded rather than replaced, and that massive capital spending now will yield dominant market positions later. They point to cash reserves, profitability, and decades of execution as proof that these aren’t fly-by-night tech firms. Meta’s AI infrastructure, for example, supports everything from content moderation to the company’s Llama models, and executives claim those assets will serve multiple generations of products. Oracle, meanwhile, insists its cloud expansion is backed by recurring enterprise contracts that justify the spending.
But Burry isn’t convinced. He believes this optimism overlooks basic economics. Hardware wears out. Software gets outdated. Energy costs rise. AI workloads demand increasingly specialized chips that can’t always be repurposed. All of that means capital expenditures will keep climbing, even as returns may start to flatten. Once that happens, margins will shrink, valuations will compress, and Wall Street’s rosy projections could unravel fast.
To anyone who remembers the early 2000s, the parallels are hard to ignore. Back then, Enron was the darling of innovation, praised for building the “future of energy.” Investors couldn’t imagine it collapsing. Then the accounting cracked, revealing how little of the company’s value was real. Today’s hyperscalers aren’t running the same kind of scam, but they may be caught in the same mindset using clever accounting to make endless growth look plausible on paper.
Burry’s warning lands at a time when markets are euphoric about AI. Nvidia’s explosive success has sent shockwaves through the industry, and every tech CEO wants a slice of that story. Capital spending on AI infrastructure has soared past half a trillion dollars globally. But if those investments are being depreciated too slowly, the true cost of the AI revolution might only become visible when the next upgrade cycle hits and the profits suddenly disappear.
For now, investors are still bullish. The Nasdaq keeps climbing, valuations remain rich, and the hype hasn’t cooled. But Burry’s message is clear: don’t mistake accounting optimism for real, sustainable profit. The AI arms race may be generating headlines and momentum, but if history teaches us anything, it’s that rapid innovation often hides financial fragility beneath the surface. Just like Enron’s promises of endless growth once captivated Wall Street, the tech titans of today might be playing a similar tune only this time, the servers hum instead of power plants. And if Burry is right again, the silence that follows could be just as loud.
