The "big short" targets AI storage again! Increasing short positions on Micron (MU.US), warning that a cyclical reversal will trigger a "violent sell-off"

According to Zhitong Finance APP, Michael Burry, famously known for shorting the US subprime mortgage market in 2008, has once again set his sights on one of the hottest tracks in the current AI wave—the storage chip sector.

According to Burry's latest trading updates disclosed to subscribers on Substack, he has further increased his short positions on Micron Technology (MU.US), Dutch AI infrastructure provider Nebius Group (NBIS.US), Semiconductor ETF - iShares (SOXX.US), and Palantir Technologies (PLTR.US), stating that the size of his short position this time is “in some size.”

This operation comes at a time when Micron's stock price continues to strengthen—on Tuesday, Micron closed up 5% at $1,096.16, extending its winning streak to the fourth consecutive trading day. Since September, the stock has risen approximately 14% and is expected to achieve its second consecutive monthly increase.

Burry has expressed clear doubts about the valuation of storage chip manufacturers. He described the share prices of these companies as having “risen to absurd levels, relative to themselves,” and predicted that once the cycle reverses, relevant stocks will face “violent selling.”

The core logic of this judgment lies in supply-side changes. Burry believes that the supply shortage of traditional DRAM over the past two years wasn’t due to a structural boom in demand, but more so a “detour” on the supply side—Samsung (SSNLF.US), SK Hynix (SKHY.US), and Micron have shifted wafer capacity on a massive scale to high bandwidth memory (HBM) required for AI data centers, thereby squeezing the supply of ordinary DDR5, pushing consumer and server memory prices higher. As these production lines gradually shift capacity back to traditional DRAM, the narrowing of the supply-demand gap Burry is betting on should follow accordingly.

In his latest trading post, Burry specifically quoted statements from Jason Chen, Chairman and CEO of Acer. Chen pointed out that while some advanced storage products are still in short supply, mature process products like DDR4 are now experiencing a “buyer’s market,” with memory inventories accumulating and capacity expansion by mainland China suppliers further suppressing price expectations.

Chen also offered a rather aggressive timeline: he expects PC prices to rise another 5% to 20% in the fourth quarter of 2026, to stabilize in the first half of 2027, and then to enter a downtrend in the latter half of 2027 as Changxin Memory’s capacity is released. He stated bluntly, “The shortage is not going to last until 2030.”

Changxin Memory’s Breakthroughs and Uncertainties

This also introduces another key piece of information from Burry’s post. Reports indicate that China’s largest DRAM manufacturer has achieved a yield rate of over 90% on its 17nm DDR5 process, narrowing the gap with Samsung’s 92%-93% yield rate for similar processes to just 2-3 percentage points. This week, Changxin Memory also announced that its fifth-generation storage chip platform has officially entered mass production. Burry commented, “If true, this is significant.”

From an industry perspective, Changxin Memory’s technology advancement is indeed accelerating. Its DDR5 products support transfer rates up to 8000Mbps, with chip density covering 16Gb and 24Gb, and actual shipments have been realized by domestic module manufacturers—for example, Sinker, a brand under Jiahe Jinwei, has had its 64GB DDR5-5600 RDIMM server memory pass tests with several major clients and has entered mass supply. However, Changxin Memory’s 24Gb density still lags about one generation behind the most advanced 32Gb DDR5 from Samsung, SK Hynix, and Micron. In terms of core patents, the company has long been constrained by the three major original manufacturers, and large-scale entry into the international supply chain still faces substantial obstacles related to compliance and intellectual property.

The biggest gap in Burry’s short thesis is that nobody can quantify when the “crossover point” will arrive. Both Changxin Memory and Yangtze Memory are expanding capacity, but neither company discloses wafer input data. The history of the storage industry repeatedly proves that capacity often comes late and is released in clusters—the DRAM downturns in 2018 and 2022 both occurred only after years of capital expenditure cycles that exceeded demand.

“Hard Data” on Fundamentals

However, Burry is shorting precisely some of the biggest direct beneficiaries of the current AI storage boom.

Micron’s third quarter for fiscal year 2026 (ending May 28) featured explosive numbers: revenue reached $41.46 billion, compared to $23.86 billion in the previous quarter and only $9.3 billion in the same period last year, representing a year-on-year increase of nearly 346%. Gross margin climbed to 84.6%, and non-GAAP gross margin reached a company record high of 84.9%. Non-GAAP earnings per share were $25.11, compared to just $1.91 in the same period last year. The company expects fourth quarter revenue to reach around $50 billion and will announce its latest financial results on September 30, Eastern Time.

These figures sharply conflict with Burry’s narrative of “absurd prices.” Micron’s profitability doesn’t stem from a valuation bubble, but from real surges in revenue and profit margins. Data center business hit a record high of $11.5 billion in single-quarter revenue, and the core data center gross margin touched 87%. Chairman, President, and CEO Sanjay Mehrotra emphasized in the earnings report that long-term strategic customer agreements will significantly enhance the persistence and predictability of the company’s financial performance.

But Burry clearly believes that such prosperity is exactly the hallmark of a cycle top. When he first established his short position on Micron in July, he stated that the AI-driven memory boom might be different from previous cycles—“but not different enough to ignore the rules of the cycle.”

Rotation and Peaks

Beyond specific short targets, Burry also has views on the broader market structure. He believes capital will start flowing out of the semiconductor sector, and that this rotation “may last longer at the beginning.” He noted, “The narrative of storage shortage and AI applications is sucking up all the market’s attention,” while at the same time, the trend of index records diverging from value stock declines “is intensifying.”

When asked whether the rotation from AI stocks to software companies would occur before a market crash, Burry’s answer was rather cautious: “Not necessarily before the crash.” He acknowledged that although market breadth is narrowing, new index highs make it difficult to maintain a bearish stance in the short term. Bull traps could indeed appear near market tops, but historically validated samples are too few, making this observation unreliable for timing the market.

This caution is not without reason. The US semiconductor sector has recently demonstrated resilience; the Philadelphia Semiconductor Index jumped 4.3% on Monday, AMD (AMD.US) surged nearly 10%, pushing its market cap past $1 trillion, while Intel (INTC.US) has risen 36% since September and is up 230% year-to-date. Micron itself rose 5.5% in a single day on September 17, and concerns about tightening storage supply are ironically serving as upward drivers for the price. In such an atmosphere, it is almost inevitable that Burry’s short position will suffer floating losses in the short term.

But looking at a longer holding timeline, Burry’s bearishness on the AI chip segment is not a spur-of-the-moment decision. Since June, he has gradually disclosed short positions on Nvidia (NVDA.US), Applied Materials (AMAT.US), Tesla (TSLA.US), Caterpillar (CAT.US), and SOXX, and officially shorted Micron in early July. In early September, he closed out Nvidia and Palantir December 2026 put options, saying the time value was decaying too quickly, though he emphasized this did not mean he had turned bullish—he still holds Palantir and Nasdaq 100 ETF - Invesco QQQ Trust (QQQ.US) 2027 put options, as well as short positions in Nvidia, Oracle (ORCL.US), Nebius, and SOXX. On the long side, he’s recently bought QXO (QXO.US), Build-A-Bear Workshop (BBW.US), Sprouts Farmers Market (SFM.US), Birkenstock (BIRK.US), and MercadoLibre (MELI.US)—these targets have almost nothing to do with the AI theme, which to some extent also reflects his avoidance attitude towards the current market mainstream.