Micron posted record results, but if memory prices keep rising like this, isn't someone losing out? I'm wondering if simply holding a semiconductor ETF is enough.

The memory boom is close to zero-sum

Micron's 87% gross margin means someone's costs rose by that much. Memory doesn't go only into AI servers. PCs, smartphones, cars, and the big tech companies building AI data centers all buy it. So these results are good news for semiconductor ETFs but a cost-increase notice for companies that buy memory to build finished products.

DRAM contract prices rose as much as 98% in the first quarter of this year, and forecasts say combined DRAM and SSD prices could rise about 130% by year-end. In that case, average PC prices would rise 17% from 2025 and smartphones 13% (TrendForce, IDC, 2026). Memory's share of PC manufacturing costs jumped from 15-18% to 35%, making it the single largest cost item (citing HP, 2026).

Winners and losers

PC makers' concerns are concrete. Dell's COO put it this way:

"We've never seen costs rise at this pace." — Jeff Clarke, Dell COO

Morgan Stanley expects global PC and hardware makers' margins to shrink by a median 60bp in 2026, while Wall Street consensus had called for a slight improvement (Morgan Stanley via Yahoo Finance, 2026). Volumes are also falling, with PC shipments expected to drop -10.4% and smartphones -8.4% this year (IDC, 2026). The 12-year low in consumer confidence released on September 30 signals that consumers have little capacity to absorb these price hikes.

Even inside semiconductor ETFs there is asymmetry

Semiconductor ETFs like SMH don't hold only memory. Companies like Nvidia, which buy memory and attach it to GPUs they sell, make up a large share. When HBM prices rise, Nvidia's costs rise too. So far AI demand has been strong enough to pass this through to GPU prices, but the bigger the memory makers' share, the more profit shifts toward memory within the same ETF. That is why Korean chip stocks with heavy memory exposure, or MU itself, move more sensitively in this cycle than SMH.

The other risk lies on the supply side. At its FQ3 results in June, Micron rose as much as +16% after hours before falling as far as -6% on news that its capex plans were larger than expected. At the time it said capex this year would exceed $25 billion and rise by more than another $10 billion in 2027 (Longport, 2026-06). Historically, oversupply has triggered memory price crashes. The end of a boom often shows up first in simultaneous capacity expansion by the three makers rather than in slowing demand. On this conference call, too, the capex figure could matter more for the stock than guidance.

Where the signals stand

As of September 29, SMH closed at 606.90, above its 10-day line (589.81), 50-day line (567.72), and 150-day line (531.48), and its entry condition (entry_ok) has been met for 5 straight sessions. It is -9.66% below its high, with 10.34% of room to the trailing stop. MU at 1,065.08 is also above all its moving averages but sits -15.12% below its high of 1,254.81. Its failure to reclaim the high even before strong results fits the community view that the market has already priced in much of the optimism.

What should investors do

• If you hold a semiconductor ETF, the trend is intact. Better to add in tranches on pullbacks near the 10-day line than during a post-earnings spike
• If you want more direct exposure to the memory upside, stocks with higher memory weight fit better than SMH. The volatility is correspondingly higher
• If you are heavily weighted in PC and smartphone hardware or low- to mid-range consumer electronics, factor in the risk of margin cuts until memory prices turn
• The numbers to watch are the capex plans of Micron, Samsung, and SK hynix, and whether quarterly DRAM contract price growth slows

Detailed charts and signals for other stocks are available on the /signals dashboard.