Will DRAM spot prices decline more than 15% from their CY2026 peak before December 31, 2026?

The committee found DRAM shortage expected to persist through 2027 based on HBM trade ratio constraints (3:1 to 4:1) and limited new capacity before ID1 wafer-out mid-2027. The question requires a >15% decline from the CY2026 PEAK within CY2026 itself — roughly 9.5 months remain. The 55-60% QoQ surge in Q1 suggests the peak may still be forming. For a 15% reversal to occur within CY2026, we would need either a demand shock (hyperscaler CapEx cut) or unexpected early supply release. The D4/LP4 EOL normalization is the most concrete near-term risk, but the committee assessed it as creating temporary shortages that normalize gradually, not a sudden 15% price break. The oligopoly's failure to prevent FY2023's 50%+ decline is relevant but that collapse occurred over multiple quarters with different supply dynamics. Within the CY2026 window, base supply constraints remain binding.

While the base case supports continued shortage through CY2026, I weight two underappreciated factors. First, the committee noted HBM demand as structural constraint is E0-E1 evidence — if AI demand plateaus even modestly, the 3:1 to 4:1 trade ratio unwinds and releases constrained commodity DRAM supply into the market. This is a binary risk with asymmetric impact on spot prices. Second, DRAM spot prices having surged 55-60% QoQ is characteristic of cycle peaks historically — the committee's own Revenue Revealer found the oligopoly did NOT prevent a 50%+ revenue drawdown. The question asks about ANY point before December 31 — that's a long tail window for a volatile commodity. Even if the base case is continued shortage, the fat tail of a demand-side shock (hyperscaler inventory correction, AI scaling plateaus) in a 9-month window pushes probability above 30%.

The resolution mechanics matter: the question measures from the CY2026 PEAK, which has not yet been established. If prices continue rising through Q2 before stabilizing, a subsequent 15% decline from a higher peak becomes easier to achieve even with modest softening. However, the committee's consensus is that the shortage persists through 2027, with new capacity not arriving until mid-2027 at earliest. The 18-36 month timeline for memory cycle downturn from capacity additions (Moat Mapper, medium likelihood) centers on 2027-2028, not CY2026. The most plausible CY2026 reversal scenario is a hyperscaler inventory correction — but with $100B+ AI CapEx commitments, this would require a significant narrative shift. I assess this as possible but below base rate for memory cycle reversals given the tight supply fundamentals documented by the committee.

The committee's finding that DRAM shortage persists through 2027 is the anchor. HBM trade ratios are actively constraining commodity DRAM supply, and new capacity (Samsung, SK Hynix expansions) does not come online until 2027. However, the 15% threshold from peak is achievable even within a broadly tight market — DRAM spot prices are notoriously volatile and a 15% pullback from an elevated peak could happen on a single quarter of inventory normalization at hyperscalers. The D4/LP4 EOL transition completing in 2-3 quarters from FQ3 FY2025 means that supply distortion clears by mid-CY2026, removing one source of tightness. Still, the fundamental supply deficit from HBM production crowding out commodity DRAM is the strongest near-term support for prices. Probability is below 50% but meaningfully above trivial.

The supply-demand fundamentals are clear: HBM production consumes 3-4x the DRAM bits per unit, structurally constraining commodity DRAM supply. No major new fabs come online until mid-to-late 2027. The committee rated the shortage as persisting through 2027 with E2 evidence and 2/2 lens agreement. For a >15% spot price decline within CY2026, you need either massive demand destruction or unexpected supply release. Neither is the base case. The D4/LP4 EOL normalization is gradual and already priced into industry planning. The China CAC ban actually reduces effective supply to non-China markets, further supporting prices. The only realistic trigger is a sharp hyperscaler CapEx pullback, which conflicts with current $100B+ AI investment trajectories. This is a fairly confident NO for CY2026, with the real risk window being 2027-2028.

I weigh the committee's unresolved debate heavily: HBM demand as structural DRAM constraint is rated E0-E1 evidence — this is the weakest link in the bull case for sustained prices. If hyperscaler orders even plateau (not crash), some of the HBM-driven constraint on commodity DRAM supply loosens. Additionally, DRAM spot markets overshoot in both directions — the 55-60% QoQ surge itself suggests the market may be overshooting to the upside, which historically precedes corrections. The committee explicitly noted that the oligopoly 'did NOT prevent 50%+ revenue drawdown' in FY2023. Still, the 9-month window and current supply deficit make a >15% reversal a minority-probability outcome. The risk is real but not yet the base case.

DRAM shortage persists through 2027 per committee findings. HBM trade ratios constrain supply. No new capacity until mid-2027. The 15% decline threshold is high for a market in severe shortage. D4/LP4 EOL creates temporary support, not headwind. The main risk is a demand shock, but AI CapEx is still accelerating. Low probability within CY2026.

Base case favors continued tightness, but memory spot prices are historically volatile. A 55-60% QoQ surge is extreme even by DRAM standards. The question measures from peak, giving maximum opportunity for a pullback to register. Even a healthy correction of 15% from an overshooting peak is plausible without fundamentals changing. D4/LP4 EOL normalization by mid-2026 removes one tightness source. Moderate but real probability.

Committee consensus is strong: shortage through 2027, HBM constraining supply, no new fabs online in CY2026. The 15% threshold from peak requires a significant event — demand shock or supply release — neither in the base case. Memory cycle downturn timeline is 18-36 months, centering on 2027-2028. CY2026 is too early for the capacity-driven reversal.