Michael Burry, who shot to fame in 2008 by shorting the U.S. subprime mortgage market, has once again turned his attention to one of the hottest sectors in this AI wave: memory chips.
Zhitong Finance APP learned that Michael Burry, who rose to fame in 2008 by shorting the U.S. subprime mortgage market, has once again turned his attention to one of the hottest sectors in this round of the AI boom—storage chips.
According to the latest trading updates Burry disclosed to his subscribers on Substack, he has further increased his position in…$Micron Technology (MU.US)$, Dutch AI infrastructure provider$NEBIUS (NBIS.US)$ 、$iShares Semiconductor ETF (SOXX.US)$and also$Palantir (PLTR.US)$short positions, adding that the short position established this time is "not small in size."
This move comes as Micron's stock continues to rally—on Tuesday, the shares closed up 5% at $1,096.16, extending their winning streak to a fourth straight session. Since the start of September, the stock has gained roughly 14%, on track for a second consecutive monthly gain.
Burry voiced clear skepticism about the valuations of memory‑chip manufacturers. He described their stock prices as having risen to "exorbitant levels relative to their intrinsic value," and predicted that, once the cycle turns, the relevant stocks would face "severe sell-offs."
The core logic behind this assessment lies in changes on the supply side. Burry argues that, over the past two years, the tight supply of conventional DRAM has not been driven by a structural boom on the demand side, but rather by supply-side "detours"—$Samsung Electronics Co., Ltd. (SSNLF.US)$ 、$SK hynix (SKHY.US)$and$Micron Technology (MU.US)$Shifting a substantial portion of wafer capacity to high-bandwidth memory (HBM) required for AI data centers has squeezed the supply of standard DDR5, driving continued price increases across both consumer‑grade and server‑grade memory. As these production lines gradually reallocate capacity back to conventional DRAM, the supply‑demand gap he had bet on is expected to narrow accordingly.
In his latest trading post, Burry highlighted remarks by Acer Chairman and CEO Jim Wong. Wong noted that certain advanced storage products remain in short supply, but for mature‑process products such as DDR4, the market has shifted to a situation where "there are more sellers than buyers," leading to accumulating inventory. Meanwhile, capacity expansions by mainland Chinese suppliers are expected to further suppress price expectations.
Chen Junsheng also laid out a rather aggressive timeline: he expects PC prices to rise by another 5% to 20% in the fourth quarter of 2026, stabilize in the first half of 2027, and then begin to decline in the mid-to-late 2027 as Changxin Memory's production capacity comes online. He stated bluntly, "The shortage cannot persist until 2030."
ChangXin Memory's Breakthroughs and Uncertainties
This also highlights another key point in Burry's post. Reports indicate that China's largest DRAM manufacturer has achieved a yield rate exceeding 90% on its 17-nanometer‑class DDR5 process, narrowing the gap with Samsung's comparable process—where yields stand at 92% to 93%—to just 2 to 3 percentage points. This week, ChangXin Memory also announced that its fifth‑generation memory chip platform has officially entered mass production. Burry commented on this, saying, "If true, that would be highly significant."
From an industry perspective, ChangXin Memory's technological advancement is indeed accelerating. Its DDR5 products support data rates of up to 8,000 MT/s and offer chip densities of 16 Gb and 24 Gb, with actual shipments already underway among domestic module manufacturers—for example, the 64 GB DDR5‑5600 RDIMM server memory under Jiahe Jinwei's Sinker brand has passed testing by several major customers and entered mass‑production supply. However, ChangXin Memory's 24 Gb density still lags behind the cutting‑edge 32 Gb DDR5 offered by Samsung, SK Hynix, and Micron by roughly one generation. Moreover, the company has long been constrained by the three major original manufacturers in terms of core patents, meaning that its full-scale entry into international supply chains continues to face substantial compliance and intellectual‑property challenges.
The biggest gap in Burry's short‑selling thesis is that no one can quantify when the "inflection point" will arrive. Both ChangXin Memory and Yangtze Memory are expanding capacity, yet neither company discloses its wafer‑start figures. Moreover, the history of the memory industry repeatedly demonstrates that capacity expansions tend to lag behind demand and are released in concentrated bursts—both the 2018 and 2022 DRAM downturns occurred only after years of capital spending outpacing demand.
"Hard data" on fundamentals
However, the very entity Burry is shorting happens to be one of the most direct beneficiaries of this round of AI‑related storage market prosperity.
Micron's fiscal third-quarter 2026 results (as of May 28) were nothing short of explosive: revenue surged to $41.46 billion, up from $23.86 billion in the prior quarter and from just $9.3 billion a year earlier, representing a year-over-year increase of nearly 346%. Gross margin climbed to 84.6%, while non-GAAP gross margin hit an all-time high of 84.9%. Non-GAAP earnings per share came in at $25.11, compared with only $1.91 in the same period last year. The company forecasts fourth-quarter revenue of approximately $50 billion and is scheduled to release its latest financial results on September 30, Eastern Time.
These figures stand in stark contrast to the "exorbitant prices" described by Burry. Micron's profitability stems not from a valuation bubble, but from a genuine surge in both revenue and profit margins. Its data center segment posted record quarterly revenue of $11.5 billion, with gross margins for its core data center business reaching as high as 87%. In the earnings report, Chairman, President, and CEO Sanjay Mehrotra emphasized that multi-year strategic customer agreements will significantly bolster the durability and predictability of the company's financial performance.
But Burry clearly believes that this boom is precisely a hallmark of the cycle's peak. When he first established his short position in Micron in July, he noted that the AI‑driven memory boom might differ from previous cycles—"but not so much as to render the cyclical patterns irrelevant."
Rotation and Top
Beyond specific short‑selling targets, Burry also offers insights into the broader market structure. He believes capital will begin to flow out of the semiconductor sector, and that this rotation "could persist for a notably extended period." He noted that "the memory shortage and the AI‑application narrative are capturing all market attention," while the divergence—where major indices hit new highs as value stocks decline—is "growing stronger."
When asked whether the rotation from AI stocks to software companies would precede a market crash, Burry's response was notably measured: "Not necessarily before the crash." He acknowledged that, despite narrowing market breadth, record-high indices make it difficult to maintain a short‑term bearish stance. Bull traps can indeed emerge near market tops, but the limited number of historically verified peak instances makes such observations an unreliable timing indicator.
This caution is by no means unwarranted. The U.S. equity semiconductor sector has continued to demonstrate resilience recently, with the Philadelphia Semiconductor Index surging 4.3% on Monday.$Advanced Micro Devices (AMD.US)$A nearly 10% surge pushed its market capitalization past $1 trillion,$Intel (INTC.US)$Since September, the stock has risen 36%, with a year-to-date gain of 230%. On September 17 alone, Micron surged 5.5%, as market concerns over tightening memory supply have instead fueled further price increases. Under these conditions, Burry's short position is almost certain to incur paper losses in the near term.
However, viewed over a longer holding period, Burry's bearish stance on the AI chip sector is far from a passing whim. Starting in June, he has progressively disclosed his positions in…$NVIDIA (NVDA.US)$、$Applied Materials (AMAT.US)$、$Tesla (TSLA.US)$、$Caterpillar (CAT.US)$and$iShares Semiconductor ETF (SOXX.US)$of short positions, he officially shorted Micron in early July. In early September, he chose to close his December 2026 put options on NVIDIA and Palantir, citing that the time value was eroding too quickly, while emphasizing that this did not signify a shift to a bullish stance; he still holds Palantir and$Invesco QQQ Trust (QQQ.US)$the 2027 put option, and retained NVIDIA,$Oracle (ORCL.US)$, short positions in Nebius and SOXX. On the long side, he recently purchased$QXO Inc (QXO.US)$ 、$Build-A-Bear Workshop (BBW.US)$、$Sprouts Farmers Market (SFM.US)$、$Birkenstock (BIRK.US)$and$MercadoLibre (MELI.US)$These holdings are hardly related to the AI theme, which to some extent also reflects his reluctance to engage with the market's prevailing narrative.