Michael Burry built his reputation by standing against the last great financial mania. Two decades on, he is standing against a new one: the memory-chip rally that has made Micron Technology one of the best-performing stocks of the artificial-intelligence boom.
Burry, the investor behind Scion Asset Management, added to short positions in Micron, Palantir Technologies, Nebius Group and the iShares Semiconductor ETF in a filing dated September 22. The disclosure shows the man who predicted the 2008 housing collapse wagering that memory prices have crested and that the companies running their factories flat-out to meet demand are priced for a cycle that cannot last.
The Scion filing is watched closely partly because of that history. Burry rose to prominence by buying credit-default swaps against subprime mortgages in the mid-2000s, a trade that paid off when the housing market collapsed and later formed the backbone of the book and film “The Big Short.” He has since disclosed a string of contrarian calls, some prescient and some not, and each new position draws investors hunting for the next crowded trade to unwind.
The reasoning leans on an unexpected source. Burry cited Jason Chen, chairman of the Taiwanese computer maker Acer, who has warned that DDR4 memory, the older standard still used in most PCs, is heading into oversupply. When the head of a PC maker, one of memory’s largest customers, sees too much supply coming, short-sellers take note.
The numbers behind the boom explain both the rise and the skepticism. Micron’s fiscal third-quarter revenue reached roughly $41.5 billion, with a gross margin near 85 percent, figures that would have seemed unthinkable for a commodity-memory maker a few years ago. The stock has climbed about 280 percent in 2026. Regular DRAM contract prices jumped 90 to 95 percent in the first quarter, according to TrendForce, then rose another 58 to 63 percent in the second, with a further 13 to 18 percent increase expected in the third.
That surge, analysts say, is not purely a story of demand. Samsung, SK Hynix and Micron, which together control about 90 percent of the DRAM market, have steered capacity toward high-bandwidth memory for AI servers and deliberately starved conventional chips, a supply-allocation decision that lifted prices across the board. The result has been the sharpest memory upcycle in decades and, in Burry’s view, the conditions for its reversal.
The memory industry’s thirty-year rule holds that booms end the way they begin, with supply. Every past upcycle has followed the same pattern: prices rise, manufacturers rush to add capacity, the new supply arrives just as demand cools, and prices collapse. Burry’s bet is essentially that the rule has not been repealed, only delayed by the speed of the AI buildout.
His position reaches beyond memory. Palantir trades at a valuation its critics have long called detached from earnings, a stock that has become a proxy for AI enthusiasm as much as for a data-software business. Its chief executive, Alex Karp, has been unusually blunt about the tensions in the market, telling CNBC this month that the leading AI laboratories may never go public because the obligations they carry are too large for public markets to hold.
Nebius, the smallest of the four targets, is in some ways the purest expression of the trade. The Amsterdam-based AI-cloud company, which emerged from the breakup of Russia’s Yandex, is spending billions to assemble clusters of AI accelerators and rent them out, a business whose fortunes rise and fall with the price of compute. If that price softens, as a memory glut would suggest, the companies paying Nebius for capacity have less reason to keep expanding. The ETF shorts, people familiar with Burry’s thinking said, are a way to bet against the entire semiconductor complex rather than any single company.
The position is small against the hundreds of billions of dollars now moving through AI markets, but its signal is outsized. Burry has been early and wrong before. He disclosed a wager against Tesla in 2020 that did not pay off at once, and his warnings in the years after the financial crisis sometimes arrived long before the market agreed. His record is uneven, which is one reason each new filing draws attention and another reason it is met with caution.
What he is really betting on is the oldest idea in the chip business. Memory is a commodity, and commodities cycle. When prices rise far enough, every producer expands, supply catches up, and prices fall, often violently. The question is whether AI has broken that pattern or merely stretched it.
The stakes are visible in the memory makers’ own behavior. The three dominant suppliers have committed to expanding production through next year, betting the shortage will persist. Burry’s filing rests on the opposite assumption: that the very expansion meant to satisfy demand will be what ends the boom, as it has in every memory cycle before this one.
For now, the market is not listening. Micron shares have shrugged off the short interest, and analysts still point to sold-out supply of high-bandwidth memory into next year as evidence the cycle has room to run. But a wager filed on a day when the memory trade looked invincible is, in the end, a bet that the top is closer than the bulls believe.
