Michael Burry is moving against one of Wall Street’s strongest stories again. The investor who became famous for anticipating the subprime mortgage crisis has increased his bearish exposure to several leading names in artificial intelligence, concentrating part of the bet on Micron Technology. The move comes just as the American memory chipmaker is going through one of the most financially favorable phases in its history.
The contrast could hardly be sharper. On one side are accelerating revenue, exceptional margins and demand for data center memory that keeps outstripping supply. On the other is Burry, convinced that these very conditions may contain the seeds of the eventual reversal. He does not necessarily dispute the growth of AI: he questions whether the current imbalance between supply and demand, and the profitability it generates, can last indefinitely.
The gap between these two readings becomes even more interesting when you look at what analysts are doing. In recent weeks, several investment houses have reaffirmed positive ratings on Micron, with price targets well above current quotes. The market is thus facing two almost opposite interpretations of the same phenomenon: for Burry, AI risks fueling the next episode of excess production capacity, while for much of Wall Street the scarcity of memory could last long enough to sustain earnings and margins.
Burry’s bet against the memory supercycle
Michael Burry’s bearish position is not limited to Micron. The investor has also increased short exposure to Palantir, Nebius and the iShares Semiconductor ETF, acting across several segments of the AI supply chain.
Micron is nonetheless a particularly interesting case because the memory industry is historically one of the most cyclical in the entire tech sector.
The mechanism is relatively simple. When demand exceeds supply, memory prices rise quickly, and with them producers’ margins and profits. But exceptional profitability encourages new investment in production. Fabs take years to build and ramp up, creating a lag between the decision to invest and the actual arrival of new supply.
When the additional capacity finally reaches the market, demand may no longer be growing at the same pace. The result can be oversupply, followed by falling prices and a violent squeeze on margins. It is a cycle the DRAM and NAND industry has gone through many times.
The question implicit in Burry’s position is therefore very precise: has artificial intelligence structurally changed this model, or is Wall Street once again extrapolating exceptionally favorable conditions into the future?
Micron’s numbers tell a completely different story, for now
The problem for anyone making the bearish case is that Micron’s current results show no signs of deterioration yet.
In fiscal third quarter 2026, the company reported revenue of $41.46 billion, compared with $23.86 billion in the previous quarter and $9.30 billion in the same quarter a year earlier.
GAAP gross margin reached 84.6%, while net income rose to $28.24 billion. Operating cash flow reached $25.39 billion.
Even more significant was the guidance for fiscal fourth quarter. Micron indicated expected revenue of around $50 billion, plus or minus $1 billion, and a gross margin close to 86%.
These are numbers hard to reconcile with the idea that the sector is already entering a phase of oversupply.
Burry’s point, however, has to be read in perspective. The issue is not necessarily what will happen in the next few quarters, but what may happen when the production investments decided during the current period of scarcity begin to turn into available capacity.
HBM, the variable that may have changed the old cycle
There is also a substantial difference from previous memory cycles: High Bandwidth Memory (HBM).
HBM has become an essential component of the accelerators used for artificial intelligence. It is more complex to produce than conventional DRAM and absorbs a great deal of manufacturing capacity.
Micron has already begun high-volume shipments of HBM4 for its main customer’s platform and has distributed samples to other customers. At the same time, it is developing HBM4E, with volume production expected in 2027.
This makes the current cycle unusual. Shifting capacity toward HBM means taking it away from the production of other memory. The expansion of AI can therefore cause very strong demand for HBM and, at the same time, lower availability of traditional DRAM.
Scarcity may thus last longer than an analysis based solely on past cycles would suggest. And this is precisely one of the arguments used by the analysts who remain positive on Micron.
Stifel: the DRAM deficit could last longer
On September 21, Stifel reaffirmed its Buy rating on Micron with a $1,500 price target.
Analyst Brian Chin expects growth to slow from the exceptional pace of recent quarters, but still considers the market to be underestimating the duration of the favorable memory cycle.
Stifel estimates that DRAM bit shipment growth could slow to 15-20% in 2027, from roughly 25-30% in 2026. Paradoxically, this slowdown in supply could be favorable for profitability.
According to Chin, truly eliminating the gap between supply and demand would require DRAM availability to grow by 40-50% in 2027.
Stifel also believes that the price per bit of HBM4 could rise significantly, further supporting margins.
Compared with Micron’s close of $1,082.28 on September 25, the $1,500 target implies upside of about 38.6%.
UBS sees demand far above supply
UBS’s target is higher still.
On September 23, Timothy Arcuri reaffirmed the Buy rating with a $1,625 price target. Compared with $1,082.28 on September 25, the gap is about 50.1%.
The rationale is particularly significant because it directly addresses the problem raised by Burry.
According to UBS’s checks, the gap between demand and supply could keep widening in 2027. Fulfillment of DRAM demand would still be around 60%, while demand for server DDR bits could grow by about 80% year over year. For servers and SSD storage, growth could even exceed 100%.
If confirmed, these numbers would make a rapid shift from scarcity to overcapacity very difficult.
TD Cowen targets $1,600
On September 16, TD Cowen reaffirmed its Buy rating and raised its price target from $1,500 to $1,600.
Using the September 25 price of $1,082.28 as a reference again, the target implies upside of about 47.8%. Here too, the valuation reflects the expectation that the strength of the memory market could last longer than traditional industry cycles.
RBC Capital, meanwhile, reaffirmed its Outperform rating with a $1,500 price target on September 18. Its upside from the September 25 price is therefore still 38.6%.
A flurry of positive ratings ahead of earnings
In recent days the number of analyst notes has risen sharply, also ahead of the quarterly results expected on September 30.
On September 23, Citi reaffirmed its Buy rating and raised its target from $1,150 to $1,300. The bank cited better-than-expected DRAM prices and a DRAM and NAND market still characterized by insufficient supply. Compared with $1,082.28 on September 25, the target implies upside of about 20.1%.
The same day, Wells Fargo maintained its Overweight rating but cut its price target from $1,525 to $1,400. Even after the cut, the target remains about 29.4% above the reference price.
Bank of America maintained a positive rating with a $1,550 target, corresponding to upside of about 43.2%. Wolfe Research set a target of $1,500, about 38.6% above the September 25 price.
Rosenblatt Securities reaffirmed its Buy rating and a $1,500 target on September 24, while on September 25 BMO Capital reiterated its positive rating with a $1,300 target, about 20.1% above the reference price.
Among the ratings published over the past month, however, one much more cautious stance stands out. Goldman Sachs maintained a Neutral stance with a $1,100 price target. Compared with $1,082.28 on September 25, the gap is just 1.6%.
The dispersion of targets is therefore considerable. It is less about the immediate direction of earnings, on which a favorable view prevails, than about the duration of the cycle.
The real risk may come from production capacity
To understand Burry’s position, you have to look beyond the next quarterly results.
Micron is investing enormous sums to expand its capacity. Samsung and SK Hynix are doing the same. As long as demand grows faster than supply, these investments are not necessarily a problem.
But the fabs being built today will produce memory in the years that follow.
The risk therefore rises if two things happen at once: a slowdown in the growth of AI investment and a substantial increase in memory production capacity. In that case, the current scarcity could quickly turn into greater availability and pressure on prices.
China complicates the picture further. CXMT (ChangXin Memory Technologies) recently announced that its fifth-generation memory platform has entered mass production. The new process would increase the number of dies per wafer by at least 50% compared with the previous generation.
Micron, Samsung and SK Hynix retain an important technological advantage, especially in the most advanced memory for AI, but an increase in Chinese output of less sophisticated DRAM could still gradually shift the global supply balance.
Burry versus Wall Street: two theses that can both be right, but at different times
This is probably the most important aspect of the whole story.
The available data do not currently show a situation of memory overproduction. If anything, they show the opposite.
Micron is posting record revenue, extraordinarily high margins and demand that continues to run up against supply constraints. The recent ratings from UBS, Stifel, TD Cowen, RBC, Citi, Bank of America, Wells Fargo, Wolfe Research, Rosenblatt and BMO largely reflect this situation.
Burry, however, is looking at a different horizon. His bet assumes that the market is attributing too structural a character to profits generated in part by an exceptional situation of scarcity. If current investments produce a sharp increase in supply just as AI demand growth begins to normalize, profitability could fall much faster than today’s expectations assume.
2027 therefore takes on particular importance.
To tell which reading is being confirmed, watching Micron’s share price day by day will not be enough. Far more telling will be the trend in DRAM and HBM prices, the growth of production capacity, inventory levels, the spending of the large data center operators and, above all, the evolution of margins.
If demand keeps outstripping supply despite the increase in capacity, the structural transformation brought about by artificial intelligence will have shown that it can extend the memory cycle beyond historical norms. If instead supply and demand begin to converge quickly, the market will have to confront the cyclical mechanism on which Burry built his bearish position.
For those following the stock, then, the most important data point in the coming quarters will not be only how much Micron’s revenue and earnings grow. It will be understanding whether the current memory scarcity is simply continuing or whether the first signs of rebalancing are appearing. Much of the contest between Michael Burry’s bearish thesis and the far more favorable one, today held by the majority of analysts, will depend on that variable.
Editor’s note
This article was originally published in Italian on money.it by Gerardo Marciano on October 03, 2026 as «Michael Burry scommette contro Micron, ma gli analisti vedono fino al +50%. Chi ha ragione?». It has been translated and adapted for an international audience by the Money.it International desk.
