The Memory Selloff Is a Leverage Event, Not a Demand Break

Positioning data says the unwind is fading. SK Hynix reports tomorrow and the Street still forecasts a record ₩64 trillion operating profit.

The memory complex is down 15% to 20% from its July highs, and every name in the basket is moving in lockstep. SNDK is down 15.3% today alone, MU 9.1%, WDC 8.4%, STX 8.3%. A China DUV milestone triggered the initial move Monday, but the scale of the liquidation — four stocks, one trade, no differentiated news — tells you what's really happening. This is a leverage unwind, not a fundamentals break.

The data supports the unwind thesis on all four dimensions: flow, positioning, gamma, and fundamentals. SK Hynix reports Q2 tomorrow, and the Street expects a record ₩64.3 trillion operating profit. The debate isn't about whether demand exists. It's about a single accounting question: how much of surging DRAM contract prices flows through to reported blended ASP.

James Thorne's Framework

James E. Thorne, chief market strategist at Wellington-Altus, called this explicitly Tuesday morning: too much leverage, too much crowding, too much exposure piled into the same trade, all of which now need to be unwound. The core distinction matters: a leverage event unwinds positioning, not thesis. A structural break would unwind earnings. The underlying AI demand story is still intact, and the supply-demand imbalance that drove memory prices sharply higher over the past year hasn't reversed. What changed is how much fast money was trapped in the move.

Positioning Says the Unwind Is Fading

MU's algo distribution score peaked at 58.3 on July 24 — the session where 48 names flagged in the broader basket. It fell to 35.8 on July 27, then 28.4 today. Three straight declines. The mechanical selling pressure is weakening, not accelerating, even as prices make new lows.

Signed flow today was essentially flat: $98 million in block buys versus $87 million in block sells, with an order-flow imbalance of +0.0001. Institutions are not panic-dumping. This is a grind driven by forced deleveraging, not organic distribution.

Gamma is amplifying every down-tick. MU's spot sits at $818, well below the gamma flip at $916. Dealers are net short gamma and selling into every decline. The $800 put wall is the next mechanical support. ATM implied volatility has spiked to 94.8%, pricing an extreme event rather than a normal repricing.

LEAPS tell the conviction story. Total call open interest is 528,000 contracts against 376,000 puts — a put/call ratio of 0.71. January 2028 upside calls in the $1,400 to $1,500 range saw net new buying today. The long-dated money is not walking away.

Fundamentals Are at Odds With the Price Action

Morgan Stanley raised Q3 DRAM ASP estimates to plus 15 to 20 percent on July 10, up from plus 3 to 8 percent, alongside HBM pricing up 8 to 13 percent and enterprise SSD up 18 to 23 percent. Korea's customs data showed memory exports at a record in June, up 57 percent quarter over quarter.

The sell-side consensus on MU remains strong buy with 42 analysts at a mean price target of $1,507 — 67 percent above today's close. No analyst has downgraded. The last rating action of any kind was KeyBanc reiterating Overweight on July 14. The market is pricing in a disaster the analysts covering the name have not penciled in.

SK Hynix Tomorrow: The Pass-Through Debate

SK Hynix reports Q2 on July 29, Seoul time. The FactSet consensus calls for revenue of ₩84.2 trillion, operating profit of ₩64.3 trillion, and net profit of ₩50.8 trillion — roughly $34.7 billion, more than seven times the year-ago quarter.

Teng Yan, head of Tessara, frames the debate precisely: Korea Investment & Securities set the bear case at blended DRAM realization of roughly +29 percent quarter over quarter, while Tessara's house view is closer to +40 percent. In Q1, contract price benchmarks rose 93 to 98 percent, while SK Hynix reported blended DRAM pricing up in the mid-60s — roughly two-thirds pass-through. The KIS bear case requires pass-through to fall toward one-half in a quarter when contracts were resetting higher. Tomorrow's print resolves it.

If blended ASP prints closer to +40 percent, the pass-through thesis survives and the short-gamma unwind could snap back violently. If it prints near +29 percent, the deleveraging has more room to run.

The China DUV catalyst that started the selloff, meanwhile, is strategically real but commercially tiny. As Lukas Frohlich (The Short Bear) detailed, five immersion DUV machines in 2026 against ASML's 130 is less than four percent of annual supply. The machines are 2006-vintage technology, not current-generation, and the real impact is on the long-term competitive landscape, not near-term memory supply. The market appears to be pricing the strategic implication, not the immediate economic impact.

What to Watch

The basket correlation on MU sits at 0.77. When the trade stops moving as one — when MU diverges from WDC or STX on an idiosyncratic catalyst — that is the signal the unwind is exhausting. SK Hynix earnings at ~9:00 a.m. Seoul time (8:00 p.m. ET tonight) is the first real test.

More research at bargo.ai/research.