A technical breakthrough by ChangXin Technology ignited the global semiconductor market within four trading sessions, pushing a crowded AI memory trade to a liquidation moment.

Goldman Sachs strategist Robert Quinn, in a latest report titled "Nasdaq Futures: Memory Crash," reconstructed the structure of this selloff through positioning data: speculative "fast money" pushed Nasdaq futures long positions to a one-year high during the week the market peaked, then was stopped out; while the approximately $70 billion in asset manager long positions has shown no signs of loosening. Goldman Sachs pointed out that this massive "slow money" position is the market's next potential source of selling.

The shockwave has left clear marks across global markets. The Goldman Sachs Memory and AI Semiconductor baskets fell 9.3% and 6.7% respectively between October 5 and 12, the Kospi index was hit hard, Kioxia shares plummeted, and major names including Samsung, SK Hynix, Micron (MU), Western Digital (WDC), and Seagate (STX) saw single-day declines ranging from 5% to 15%.

A trade with triple-digit gains is starting to crack

Goldman Sachs memory basket data outlines the full arc of this trade: the basket index started at around 600 points a year ago, climbed to a peak of approximately 2,675 points by late June this year, and has now pulled back to around 1,900.

This means that even after this round of adjustment, the trade still carries book gains of more than three times compared to a year ago. But measured from the peak, it has retraced about 25% from the June high. In such a crowded trade, a 25% drawdown means the distance between victory and a margin call has narrowed dramatically.

The direct trigger for this selloff was ChangXin Technology's claim of "key progress" on a new 4F² DRAM architecture, asserting it could achieve higher memory density ahead of Samsung and SK Hynix—though according to Bloomberg, this claim has not yet been verified. Tesla CEO Elon Musk immediately stated on social media that "when China figures out how to make a product, its scaling capacity often exceeds the combined total of the rest of the world," and reminded the market that "three years ago memory was just a low-margin commodity." These remarks further intensified market concerns about whether memory's extraordinary profits can be sustained.

Leveraged funds: Entered at the top, then turned

Goldman Sachs' second finding comes from the cost-of-funds market. Quinn noted that the premium of the Nasdaq 3-month funding rate relative to the federal funds rate continued to climb during the memory stock decline, indicating that leveraged longs initially did not retreat, before a clear turn emerged on October 12 when the market weakened.

Currently, the premium of the Nasdaq 3-month funding rate over the federal funds rate is approximately 76 basis points, roughly double the March low, but still below the peak level of about 124 basis points seen at the June market top.

The Nasdaq's funding premium relative to the S&P 500 provides an even clearer timeline: after hitting a year-to-date high of about 7 basis points in early October, the premium turned sharply lower during Monday's selloff. Leveraged funds completed their final round of adding positions right at the top.

Fast money: Built positions at a one-year high, caught the peak

The most critical chart in Quinn's report comes from the CFTC's Commitments of Traders (COT) report.

Data shows that the total long positions of "other" category traders in the Nasdaq futures market (typically corresponding to fast money, proprietary trading desks, and other speculative funds) reached approximately $9.6 billion on October 6, more than tripling from about $3 billion in early July. Their share of open interest also rose from about 1.4% at the end of June to about 5.4%, both the highest levels in the past year.

October 6 happened to be the week before OpenAI's revenue forecast triggered market concerns, and five days before ChangXin Technology's announcement. Fast money pushed long positions to a one-year peak during the week the market reversed, then was stopped out—highly consistent with fast money's habitual behavior pattern. Quinn, using the wording "may have reduced recently established long positions," tactfully expressed the same judgment: this batch of longs has likely already exited, and did not successfully escape at the highs.

Slow money: $70 billion in longs, completely unmoved

However, what truly unsettles market bulls is Goldman Sachs' final finding.

As of October 6, asset managers' total long positions in Nasdaq futures stood at approximately $70 billion, nearly matching the record of about $71 billion set in June, and up sharply from about $44 billion at the end of March. Quinn noted that this batch of asset manager long positions shows no significant correlation with memory and AI semiconductor price action—in other words, they have not adjusted positions due to the memory stock decline.

This batch of capital has logical support for entering: Goldman Sachs had estimated that Micron alone would contribute 19% of S&P 500 earnings growth in the third quarter, and the memory industry's margins are extremely lucrative—Nanya Technology's latest quarterly gross margin reached 85.7%. But according to Bloomberg, Jon Withaar of Pictet Asset Management pointed out that memory stocks "have become value stocks, cheap on current earnings," and "retail investors and fast-money hedge funds don't chase value stocks." Fast money's actions have already confirmed this judgment.

The question is whether the long-term capital holding approximately $70 billion in longs will view a 25% drawdown as a buying opportunity, or begin to re-examine the core question raised by Goldman Sachs' Korea team ahead of Samsung's October 29 earnings call: can memory supply remain tight through 2028—by which time Goldman Sachs expects ChangXin Technology's DRAM supply to reach half of SK Hynix's.

The next test: Slow money's patience

Quinn remains cautious about asset managers' moves in the report, calling their impact "still debatable," and noting that Goldman Sachs' proprietary trading desk expects buying to emerge in the semiconductor sector's decline, with ASML and TSMC (TSMC) earnings being the first near-term checkpoint.

But the positioning picture Goldman Sachs paints reveals a structure familiar to market participants: leveraged funds completed their final position-building at the top, fast money reached a one-year high in positioning during the week the market reversed, and the largest long position—approximately $70 billion in asset manager positions—remains near historic highs even as the underlying trade has retraced 25% from its peak. Meanwhile, a Chinese competitor with at least narrative credibility is threatening the memory industry's margin story, and Korea's market—a leading indicator—has already fallen more than 30% from its high.

Fast money is the canary in the coal mine; asset managers are the mine itself. The direction of the next phase may have little to do with ChangXin Technology's technical progress, and more to do with how long slow money's patience can hold—and when a "value" trade keeps declining, patience is often the first thing to run out.