SemiconductorsFEATURE

DRAM Prices Nearly Doubled in a Quarter: Why This Memory Supercycle Is Different

E
EffectStory 編輯部Editorial Team
Published · Updated
General-purpose DRAM contract prices jumped roughly 93–98% quarter-on-quarter in Q1 2026, with NAND Flash close behind. The surge stems from memory makers diverting capacity to HBM and AI servers under long-term agreements, leaving standard DRAM supply unable to catch up. Micron's 2026 HBM output is already sold out through 2027, and record revenue at SK Hynix and Micron shows the demand is structural, not a typical inventory-driven cycle.

How steep is the surge in DRAM and NAND prices?

General-purpose (PC and server) DRAM contract prices rose approximately 93–98% quarter-on-quarter in the first quarter of 2026CITE:E1. A further increase of roughly 58–63% quarter-on-quarter is projected for the second quarterCITE:E1. NAND Flash contract prices are set to rise about 70–75% quarter-on-quarter in the same second quarter, showing the rally spans both DRAM and NAND rather than a single product categoryCITE:E3.

MetricPeriodChange (QoQ, contract price)
General-purpose DRAMQ1 2026 (realized)+93% to +98%
General-purpose DRAMQ2 2026 (estimate)+58% to +63%
NAND FlashQ2 2026 (estimate)+70% to +75%

Why are memory prices spiking, and how has AI crowded out standard DRAM capacity?

Memory makers have redirected production capacity toward HBM and server-grade memory, locking in a portion of supply through long-term agreements (LTAs)CITE:E2. Because new memory fabrication capacity takes a long time to build, supply cannot respond quickly to this demand shift, leaving standard DRAM output constrained in the near termCITE:E2.

Why has HBM supply become the tightest bottleneck, and how are makers responding?

Micron has stated that its entire calendar-year 2026 HBM supply — both price and volume — is already sold out, with demand booked into 2027CITE:E4. On the technology side, HBM generations continue to advance: HBM3E delivers roughly 1.2 TB/s of bandwidth per stack, while HBM4 — standardized by JEDEC in 2025 — reaches 2 TB/s per stackCITE:E6. All three major memory makers are moving into HBM4 production during 2026CITE:E6.

HBM generationBandwidth per stack
HBM3E~1.2 TB/s
HBM4 (JEDEC 2025 standard)2 TB/s

How has AI chip demand pushed memory makers to record results?

SK Hynix and Micron both posted record results driven by HBM and AI server demand. SK Hynix reported second-quarter 2026 revenue up 257% year-on-yearCITE:E5. Micron posted fiscal third-quarter revenue of $41.46 billion, a record figure, up approximately 346% year-on-yearCITE:E5.

CompanyMetricResult
SK HynixQ2 2026 revenue, YoY change+257%
MicronFQ3 revenue$41.46 billion (+346% YoY)

Why is this "memory supercycle" different from past cycles?

SK Hynix frames the current upcycle as structurally different from prior memory cycles: past cycles were driven mainly by producers cutting output to clear excess inventory, while this one is driven by a structural increase in the amount of memory required per unit of AI computingCITE:E7. The 2022–2023 downturn led producers to cut capital expenditure sharply, and new memory capacity now takes 18 to 24 months to come online, limiting how quickly supply can catch upCITE:E7.

What this means

The numbers line up into a single constraint: Micron's HBM supply is already booked through 2027CITE:E4, all three major makers are only now transitioning into HBM4 productionCITE:E6, and new fab capacity takes 18 to 24 months to addCITE:E7 — a timeline that cannot move as fast as quarter-on-quarter DRAM prices that have already risen by as much as 98%CITE:E1. The record revenue at SK Hynix and Micron is the direct financial readout of that same capacity gapCITE:E5.

📊 Evidence

FAQ

How steep is the surge in DRAM and NAND prices?

General-purpose (PC and server) DRAM contract prices rose approximately 93–98% quarter-on-quarter in the first quarter of 2026CITE:E1.

Why are memory prices spiking, and how has AI crowded out standard DRAM capacity?

Memory makers have redirected production capacity toward HBM and server-grade memory, locking in a portion of supply through long-term agreements (LTAs)CITE:E2.

Why has HBM supply become the tightest bottleneck, and how are makers responding?

Micron has stated that its entire calendar-year 2026 HBM supply — both price and volume — is already sold out, with demand booked into 2027CITE:E4.

How has AI chip demand pushed memory makers to record results?

SK Hynix and Micron both posted record results driven by HBM and AI server demand. SK Hynix reported second-quarter 2026 revenue up 257% year-on-yearCITE:E5.

📎 Sources

  1. trendforce.com
  2. trendforce.com
  3. trendforce.com
  4. futurumgroup.com
  5. effectstory.com
  6. effectstory.com
  7. news.skhynix.com

Related data

Author's TakeEffectStory 編輯部

The clearest signal in this cycle isn't the price jump itself — it's the sold-out status behind it: Micron's entire 2026 HBM supply is already booked through 2027, while all three major makers are only just moving into HBM4 production. That timing gap matters, because new memory capacity takes 18 to 24 months to come online, a lag that predates today's demand rather than following from it. The metric worth tracking next is how fast HBM4 output actually ramps at the three makers, since that — not the current DRAM contract price — determines when this capacity constraint eases. SK Hynix's 257% revenue growth and Micron's record $41.46 billion quarter aren't isolated results; they're the direct readout of the same supply squeeze.

E
EffectStory 編輯部Editorial Team

Related

FEATURE

AI Capex Nears $725 Billion in 2026: Rational Arms Race or Bubble?

Microsoft, Alphabet, Amazon, and Meta are on pace for roughly $725 billion in combined 2026 capital spending, up 77% from 2025, and NVIDIA's data center revenue confirms real near-term chip demand. But Bain projects an $800 billion annual revenue gap by 2030, investor Michael Burry alleges depreciation-driven profit inflation at Oracle and Meta, and a NVIDIA-OpenAI-Oracle financing loop raises structural questions even as bulls cite self-funded cash flow and scarce data center capacity.

EffectStory 編輯部 ·
FEATURE

Sovereign AI: How Gulf States, Nordic Funds, and Private Capital Are Racing to Own the World's Compute

Saudi Arabia's HUMAIN, the UAE's G42-led Stargate UAE, and Qatar's QIA have each committed multi-billion-dollar deals to build national AI compute, led by HUMAIN's $10 billion AMD partnership and a plan for up to 600,000 NVIDIA GPUs. But US export approval currently caps HUMAIN's chip access at roughly 35,000 GB300-class units, while Norway's NBIM and SoftBank show that not all "sovereign AI" exposure means building a data center.

EffectStory 編輯部 ·
BRIEF

SK Hynix Eyes ₩10 Trillion Japan Plant as Semiconductor ETFs Rally

SK Hynix (SK海力士) is planning a ₩10 trillion plant investment in Japan, a move that coincided with gains in two Japan-focused semiconductor ETFs. The plan follows a first half of 2025 in which the company posted ₩39.87 trillion in revenue, ₩15.10 trillion in net profit, and a debt-to-equity ratio that fell to 48.13% from 62.15% at the end of 2024.

EffectStory 編輯部 ·