The bounce came with names attached. Micron’s was the smallest. Friday is why.

The bounce, ranked — and what the ranking says

Monday’s tape was a dip-buyer’s tape: the Nasdaq up well over a point by late morning, the Dow briefly above 53,000 for the first time ever, and semis leading. The bids had reasons attached. Goldman raised AMD to a $640 target from $450 — AMD ripped more than 8%. Broadcom extended its custom-chip partnership with Apple through 2031 — AVGO jumped. And memory got the full defense: the UBS forecast raise, the BofA “healthy reset” note pointing out that memory now carries 35–40% of cloud AI capex yet “trades at sub-par 10x forward PE,” Citi’s catalyst watch, and Korea’s Friday session — while US markets were closed for the holiday — already up big, with the Kospi +5.8% and SK hynix +10.9% on the Samsung price-hike reports.

Rank Monday’s bounce against last week’s fall, though, and one bar sticks out. Every storage name that fell 15–23% bounced 4–7.4%. Micron fell right with them — and bounced 3.7%, less than SanDisk, less than Seagate, less than Western Digital, less than Marvell, barely more than the SMH index itself. The names defending Micron by name — UBS, BofA, Citi — were defending the whole complex; the market bought the complex and still under-bought Micron.

Chart 2 — The fall and the bounce, side by side

Everything that fell got bought back Monday — Micron least among memory. The overhang has a date on it.

Upper bars: change in closing price, June 30 → July 2. Lower bars: July 2 close → July 6 ~noon ET. Micron in navy; ASML — a new position, initiation note upcoming — in purple. Sources: bpleon price feed; analyst actions per Yahoo Finance/CNBC reporting, July 6.

Note Nvidia again, the control group in both directions: −2.6% through the unwind, +0.8% on the rebound. It never priced a demand problem, so it had nothing to un-price. The violence, both ways, lived in memory and the things priced like memory — which is what you’d expect if the move was positioning being forced out and then walked back in, not the cycle ending.

Chart 3 — The fork since the print

Two weeks after the best quarter in memory history: the market is at highs, and Micron is still below its own blowout.

Closing prices indexed to the June 24, 2026 pre-print close (=100); the July 6 point (dashed leg) is a ~noon ET snapshot. Micron remains the only line meaningfully below 100 — the bounce has not yet repaired the round trip. Source: bpleon price feed.

Tuesday re-anchors the earnings. Friday clears the overhang.

Tuesday: Samsung’s preliminary Q2. Consensus sits near ₩86 trillion of operating profit — roughly $56 billion, about 18× a year ago, a third consecutive record — on memory pricing that Korean brokers estimate rose ~55% quarter-over-quarter in DRAM and ~60% in NAND. A Samsung Securities strategist framed the moment better than we could: “The market now stands at an inflection point where doubts about an AI peak-out must be resolved with numbers.” That is the print’s job. A blowout re-anchors the whole complex’s earnings power days before the listing; a miss — after this specific week — would give the unwind its second leg.

Friday: the SKHY debut. The structure is now official: roughly $28 billion at current prices — 17.79 million new shares, about 2.5% of the company, split ten-to-one into ADRs — the largest ADR offering on record, ahead of Alibaba’s 2014 debut. Two things are true about it at once. For Micron’s multiple, it is the end of an era: the only-US-listed-memory scarcity premium we cut our target for last week dies at the open, and Monday’s lagging bounce is the market pricing that in real time. For the industry, it is the supply-side question made flesh: the proceeds are earmarked for new Korean fab capacity and EUV lithography tools — the capital that eventually answers this shortage. How SKHY trades, and how memory trades around it, is the cleanest test we will get of whether last week was positioning (our read, and now UBS’s and BofA’s) or the beginning of the cycle question.

What the defense misses, and what the panic missed, is the same thing: the pricing data was never the problem. TrendForce’s July 3 update has third-quarter DRAM contract prices rising another 13–18% — slower than Q2’s ~58–63% and Q1’s ~90–95%, which is the deceleration we flagged, but still rising, with spot prices drifting sideways-to-up through the crash week. The bear case last week wasn’t priced off shipping data; it was priced off headlines — Meta’s excess-compute cloud, an Anthropic–Samsung custom-chip report, Apple exploring Chinese memory suppliers, a short-seller’s position — every one of them a demand-side narrative landing on the most crowded positioning of the half. Narratives reprice multiples. Contracts pay earnings. Both are real; they are not the same thing.

The book: the runner, the stop, and a new name

Nothing about the plan changed, because nothing about the plan was supposed to change on a bounce we said could come. The ~40% Micron runner stays behind the pre-committed line: a close below $900 and we are out in full — at midday that line sits 11.0% below. We add nothing before Friday’s listing clears and nothing before the late-July hyperscaler capex prints referee the demand side, July 22–31. The base target stays ~$1,300, the multiple-cut re-rate we published Thursday — and we’d note the Street’s Monday defense brackets it usefully: BofA’s $1,550 sits above us on the same earnings, and UBS’s undersupplied-to-2028 call is the bull case for the EPS inputs we deliberately left unchanged. If Samsung’s print and the listing both clear clean, the path back toward the bull cells runs through those two facts, not through Monday’s bounce.

Why ASML, and why now

The unwind gave us the entry we’d been waiting for in the one name that doesn’t have to answer the question the rest of the complex is fighting about. ASML fell 11.1% across the two-day rout — sympathy damage in a name whose order book runs on multi-year lead times, days after a June in which its Amsterdam shares set a record close. Here is the asymmetry we’re buying, and it was on display this week: if memory stays tight, the shortage argument (UBS: undersupplied until at least 2Q28) forces capacity expansion — and capacity is bought from ASML. If supply answers — the bear case — it answers through ASML: SK hynix’s $28 billion raise is earmarked for new fabs and EUV tools, and Samsung’s record profits fund the same shopping list. The toolmaker gets paid by both sides of the memory-cycle question. That is the EUV-bottleneck logic our June 29 outlook sketched, converted into a position. It reports Q2 on July 15, having already raised full-year guidance to €36–40 billion in April. The full initiation — thesis, numbers, kill-shots, sizing — is the next piece in the queue.