Momentum stocks like Micron fall into correction territory. Is it something bigger or time to buy?

Chip stocks sold off more than 10% via the MTUM momentum ETF from its June 22 high (then rebounded) as traders framed the move as an orderly momentum unwind rather than collapsing AI demand. UBS/Goldman/DWS pointed to intact “memory fundamentals” and emphasized aggressive profit-taking and momentum outflows, while JPMorgan suggested it may be time to “buy the dip.” Memory was pressured by Samsung’s revenue miss versus sky-high expectations and Amazon’s note that it won’t issue more debt this year, and further catalysts loom with SK Hynix’s planned listing of 18M U.S. shares to raise nearly $30B and an upcoming earnings hurdle.

Analysis

This looks more like a factor unwind than a demand shock, but that distinction matters: when momentum de-grosses, semis can trade on flow rather than fundamentals for several sessions even if the AI capex story is intact. The near-term winner is relative quality inside the group — names with diversified revenue streams and pricing power should reassert faster than pure beta exposure, while the most crowded momentum baskets remain vulnerable to another 3-5% air pocket if systematic sellers keep reducing risk.

The next 1-3 months are about whether “good enough” fundamentals can still beat a bar that has been lifted by crowded positioning. Memory is the most fragile part of the chain because it is the easiest place for the market to extrapolate too much too fast; any sign that hyperscaler spending is being funded with more discipline, or that inventory is building ahead of earnings, will hit MU/SSNLF-style exposure before it hits the broader AI complex. The Hynix share sale is a useful tell: if the market absorbs that supply cleanly, this turns into a tradable reset; if not, it signals that investors are demanding a discount for future capex overhang.

Contrarian risk: the consensus is probably underestimating how long valuation compression can persist even when the thesis is right. “AI is fine” does not prevent a 10-15% drawdown in the segment if the market decides the monetization curve is still lagging the infrastructure curve. Over 6-18 months, the structural winners remain the companies that monetize picks-and-shovels spend with recurring software/networking economics, but in the meantime anything tied to financing-sensitive capex or memory supply can lag until breadth improves and the earnings bar comes down.

Key Decisions for Investors

• Long AVGO into weakness; use a 1-3 month horizon and only add on pullbacks rather than strength. Base case is that it outperforms the broader chip basket as the market rotates from crowded beta into quality AI infrastructure. Falsify if AVGO loses relative strength vs SOXX on a post-earnings basis.
• Pair trade: long AVGO / short MTUM for a 2-6 week factor-neutral expression. This isolates the unwind of momentum flows while keeping exposure to AI infrastructure intact. Risk is a sharp re-acceleration in factor performance; cover if MTUM reclaims its prior high on strong breadth.
• Stay tactically underweight memory-linked exposure for the next 1-3 weeks, especially MU and SSNLF proxies, into the Hynix share supply event and the start of earnings season. If the listing is absorbed with weak tape or if memory guidance comes in merely in-line, expect another leg of multiple compression.
• Do not add to AMZN on the assumption that capex will re-accelerate from debt markets; the more important signal is whether hyperscaler spending is being rationed by cash-flow discipline. Revisit after the next cloud capex commentary; if capex growth slows, suppliers with the most Amazon exposure get hit first.
• If SOXX/SMH holds the recent low and breadth broadens over the next 5-10 trading days, consider a call spread instead of outright longs to capture the rebound while limiting downside to a renewed momentum flush.