The June correction may have been deleveraging, or it may mark a peak in expectations. A falling share price alone does not prove that DRAM pricing and profits have reversed.
A shortage supports profits; it does not set the share price
Base 2027 demand is 47.41 EB versus Wu Zihao's 45.12 EB supply forecast, leaving a 2.29 EB current-year gap. That supports pricing and utilization, but equities discount changes in future earnings.
Supply-demand becomes a testable peak signal only when price momentum, vendor ASPs, EPS revisions and valuation weaken together.
The two stocks need not peak together
SK hynix has greater exposure to HBM and data centers, so changes in AI order growth tend to enter its valuation earlier. Micron is more diversified but more sensitive to conventional DRAM pricing and earnings revisions.
Our base case places SK hynix's high-risk window in Q4 2026 to Q1 2027 and Micron's in Q1 to Q2 2027. These are conditional windows, not fixed dates.
What would invalidate the view
Risk moves to red if DDR5 and server DRAM momentum stays negative for four weeks, 2027 supply is revised materially higher and forward EPS stops rising.
If Rubin and ASIC shipments keep being revised up while supply remains constrained, the peak window could shift into the second half of 2027.
For industry research only. This is not investment advice. Forecasts depend on assumptions and may differ materially from actual results.
Methodology