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

The Q2 data presents a nuanced picture for DRAM spot pricing. On one hand, supply tightness has intensified — DRAM prices surged mid-60s% sequentially (accelerating from +20% in FQ1), management sees supply constraints persisting 'beyond calendar 2026,' PC/smartphone OEMs face low double-digit unit declines from allocation pressure, and NAND cleanroom space is being redirected from NAND to DRAM (further constraining NAND but also redirecting investment away from potential DRAM capacity additions). On the other hand, several cycle-peak warning signals emerged: (1) nearly all DRAM revenue growth is pricing-driven with only mid-single-digit bit shipment growth — the textbook pattern for cycle peaks, (2) the stock declined 4.8% post-earnings despite massive beats — the market may be pricing imminent reversal, (3) a new 'Pricing Reversal Trap' scenario was identified at 15-25% probability over 24 months, and (4) FY2026-2027 CapEx of $60-65B+ across the industry is the mechanism for eventual supply normalization. The question is whether normalization begins before December 2026 — given 9 months remaining and no new capacity coming online until mid-2027+, a >15% spot decline requires a demand shock rather than supply normalization.

I weight the cycle-peak indicators more heavily. DRAM prices have now surged cumulatively ~100% over two quarters (20% FQ1 + mid-60s% FQ2). The historical DRAM cycle shows that the steeper the ascent, the more violent the eventual reversal. The market's 4.8% sell-off on massive beats is a signal — sophisticated investors may see the same cycle-peak pattern we do. The 'Pricing Reversal Trap' was identified at 15-25% probability over 24 months; within the 9-month CY2026 window, this translates to roughly 8-15%. However, I add additional probability for demand-shock scenarios: hyperscaler CapEx has been escalating unsustainably, and any deceleration signal could cause inventory destocking that amplifies a price decline. The OEM unit decline data (low double-digit) could also be a leading indicator — if OEMs are reducing production due to allocation, they are building less finished goods inventory, which means a supply normalization could hit an undersupplied end market where OEMs have already adjusted down. The reversal, when it comes, could be sharp.

The balance of evidence tilts toward continued tightness through CY2026 but with non-trivial reversal risk. Key arguments for sustained prices: (1) HBM trade ratios structurally constrain commodity DRAM, (2) no new capacity until mid-2027, (3) NAND-to-DRAM cleanroom conversions tighten DRAM supply further. Key arguments for >15% decline: (1) DRAM spot markets can move 15% in weeks during sentiment shifts, (2) inventory correction risk if hyperscalers pre-ordered aggressively, (3) 75% quarterly repricing means contract prices can quickly reflect spot declines, amplifying any weakness. The critical distinction is that a >15% spot decline from the CY2026 PEAK only requires prices to peak in Q1-Q2 and then partially retrace in H2 — it does not require prices to fall below current levels, just below 85% of whatever the peak turns out to be. Given the extraordinary price acceleration, a Q3-Q4 normalization that retraces part of the H1 gains is plausible.

The Q2 earnings data paradoxically both supports continued tightness AND raises cycle-peak concerns. The supply fundamentals are unambiguous: management says tightness persists beyond CY2026, OEMs face double-digit unit declines, NAND space redirected to DRAM. But the demand structure is concerning: price-driven growth with flat volumes is how every DRAM cycle has peaked historically. The question resolution hinges on timing — can a >15% spot decline from peak occur within the ~9 months remaining in CY2026? Three scenarios could trigger it: (1) hyperscaler inventory correction in H2 2026 — AI CapEx is the fastest-growing demand segment and the most susceptible to sudden rebalancing, (2) China demand weakening from geopolitical escalation — the $6-7B addressable market gap could widen, (3) Samsung aggressive pricing to regain share after falling behind in HBM. None of these are base case, but collectively they represent material probability.

I take the most bearish view in the ensemble because I believe the market is underweighting two factors. First, the 'Pricing Reversal Trap' scenario at 15-25% over 24 months maps to roughly 10-18% over 9 months — and that scenario explicitly requires only supply normalization, not a demand shock. With $60-65B+ in industry CapEx for FY2026-2027, some supply acceleration is possible even before new fabs come online (capacity optimization at existing fabs, yield improvements, reduced HBM allocation if AI demand plateaus). Second, DRAM spot prices are a leading indicator that can decline sharply even when fundamentals remain tight — spot represents a small fraction of total volume, and sentiment shifts can create spot price volatility that exceeds fundamental reality. A 15% spot decline from peak could occur from sentiment alone while contract markets remain tight. The Q2 stock reaction supports this view — the market sees the same cycle-peak signals.

The supply constraints are genuinely severe — management's language has strengthened from 'medium-term' to 'beyond calendar 2026,' OEMs face production cuts, and NAND cleanroom space is being redirected to DRAM. These are not the conditions that precede a >15% spot price decline. The new capacity that would normalize supply does not come online until mid-2027+. A demand shock (hyperscaler CapEx cut, geopolitical disruption) could trigger a spot price decline, but the AI CapEx cycle shows no sign of slowing — if anything, it is accelerating with NVIDIA Vera Rubin, Google's continued TPU investment, and AWS Trinium ramping. I reduce slightly from the prior 30% because the Q2 data strengthens the supply constraint narrative.

Supply tightness persisting beyond CY2026 per management, OEM unit declines from allocation, NAND-to-DRAM cleanroom shifts all support continued pricing. But price-only growth (flat bits) is cycle-peak pattern and spot markets can move 15% quickly on sentiment. Balance of evidence: slightly below prior 30%.

The ~100% cumulative DRAM price surge over two quarters is historically extreme. While supply constraints are real, the magnitude of the price increase creates a higher base from which a 15% decline is easier to achieve. The question measures from the peak — if prices plateau or slightly decline in H2, the peak was H1 and a partial retracement is possible. 4.8% stock decline on massive beat signals the market sees peak cycle risk.

No new DRAM capacity online until mid-2027. Supply constraints are the most severe in years — OEMs cutting production volumes. HBM trade ratio continues constraining commodity DRAM. A >15% decline requires a demand shock (hyperscaler CapEx freeze) or sentiment-driven spot sell-off. Neither is base case. Prior 30% was too high given the strengthened supply narrative.