Memory makers mint megabucks while buyers beg for mercy

Counterpoint Research reported that Q3 2026 DRAM revenue surged 369% year over year to nearly $200 billion, while bit shipments rose only about 4%; average selling prices reached $1.90 per gigabit. AI and datacenter demand are keeping supply tight, with prices expected to remain high and shortages forecast to last until at least 2030. Samsung led with 38% of DRAM revenue, while the article also cited strong profit figures for SK hynix and Micron; higher memory costs are weighing on PC and smartphone shipments and buyers.

Analysis

The key market signal is pricing power, not unit growth: suppliers appear able to monetize scarce capacity while downstream device makers absorb the squeeze. That supports near-term memory earnings, but it also shifts risk toward demand destruction, weaker low-end device mix, and customer efforts to redesign around scarce or expensive memory. LTA coverage may cushion near-term pricing, yet can create divergent supplier results depending on contract terms and product mix; it does not guarantee the same earnings trajectory for each maker.

For the next 1–3 months, watch supplier guidance, contract pricing, HBM mix and shipment growth—not just reported revenue. For 6–18 months, the thesis is vulnerable if AI infrastructure spending slows, memory efficiency reduces content per system, or new capacity arrives earlier than expected. The claim that tightness lasts to 2030 is a forecast, not a basis for underwriting that duration. Consumer-device weakness could also feed back into orders and inventory, even if datacenter demand remains firm.

Contrarian risk: the scarcity narrative may be extrapolating peak-cycle economics. The article’s stated profit figures and periods appear difficult to reconcile across companies and scopes; verify filings, units and accounting periods before using them in estimates. No valuation or consensus data is supplied, so the case supports a conditional relative-value setup, not an unconditional chase.

Key Decisions for Investors

• Prefer a conditional relative trade: long SKHY versus MU only if the next reports confirm SKHY’s relative HBM/memory mix or share gains and sustained pricing. Reassess or close if that relative advantage reverses; do not infer valuation upside from the article alone.
• For the next quarter, track DRAM/HBM contract pricing, supplier bit shipments, LTA disclosures and inventory at PC and smartphone makers. A price rise without resilient shipment trends would weaken the earnings-quality thesis.
• Avoid treating a shortage through 2030 as established. Reduce bullish exposure if AI-capex guidance weakens, memory prices roll over, or suppliers signal capacity arriving sooner than expected.
• Verify the article’s company profit figures against audited or reported filings before updating forecasts; mismatched periods, units or business scopes could materially distort apparent earnings momentum.