The storage segment long-term AI bull market has not ended, but short-term volatility and a re-evaluation window have opened

Overview

This report discusses the three storage themes currently most debated by investors: first, whether AI overhang means excess compute when the main AI capex participants can deploy and sell compute; second, why related stocks did not see a clear valuation re-rating after the long-term agreement (LTA) was announced; third, whether the current peak in the change rate of the storage cycle is a cycle high point or a healthy reset within an extended AI-driven cycle. The conclusion is that the long-term structural logic remains, but short-term stock performance may come under pressure before the earnings season.

Core views

The core view is that the storage industry remains cyclical, and price year-over-year change, inventory, and earnings revision breadth are approaching peak rates of change, so the market may move from strong momentum into a range-bound phase or correction; this does not mean the cycle has ended. AI capex, Agentic AI, and token penetration still support long-term demand. If hyperscalers maintain or increase capex, a near-term pullback may become a better entry point for storage stocks; if they cut capex, the narrative of compute overhang may persist.

Analysis framework

The report combines investor controversy tracking, cyclical momentum assessment, crowded positioning monitoring, AI capex chain analysis, and company valuation framework, with emphasis on hyperscaler earnings and guidance, DRAM/NAND price expectations, revision breadth, and the market’s valuation response to LTA.

Methodology notes

• Valuation methodsResidual income valuation model

  Residual income valuation model

  The report discloses that the base-case valuation target is based on a residual income model, assuming cost of equity of 11.5%, risk-free rate of 5%, risk premium of 6.5%, beta of 1.0, and perpetuity growth rate of 3%.

• cycle analysisPeak rate of change

  peak in rate of change

  The report believes memory prices versus year-over-year, inventories, and breadth of earnings revisions are nearing the upper end, and stocks usually react before fundamentals fully reach the cyclical peak.

• theme analysisAI capex and token demand framework

  AI capex and token demand framework

  The report links memory demand to hyperscaler capex, Agentic AI, token maxing/token minimization, low-cost open-source models, and the monetization ability of AI infrastructure.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• DRAM

  One of the most favored storage subsectors, and a key beneficiary of AI capex and bottleneck formation.

  Strengths

  Strong earnings revisions, high AI demand sensitivity, and the market pays closer attention to price hikes and supply-demand tightness.

  Weaknesses

  Year-over-year price change and earnings revision breadth are already near historical highs, so short-term momentum may slow.

  Comparison

  Compared with NAND and memory module makers, DRAM better aligns with capital flow preferences and bottleneck logic.

  Risks

  Hyperscaler capex cuts, compute overhang narrative, crowded-position unwind, and peak in price momentum.

• legacy memory

  The report explicitly states a preference for legacy memory.

  Strengths

  Benefits from capital allocation and potential supply bottlenecks, and may show better relative performance within the AI value chain expansion.

  Weaknesses

  Still exposed to storage cycle swings and end-demand volatility.

  Comparison

  More favored than memory module makers, while relative thematic elasticity is more dispersed than DRAM.

  Risks

  Demand slowdown, elevated inventories, competition intensification, and weakening price momentum.

• NAND

  The report is relatively less favorable toward NAND.

  Strengths

  Some subsegments such as enterprise SSD still have upside in price expectations.

  Weaknesses

  Preference ranking is lower than DRAM and legacy memory, and 4Q26E projected price growth is relatively limited.

  Comparison

  NAND has weaker relative appeal than DRAM in this report.

  Risks

  Terminal demand below expectations, high inventories, and decelerating price gains.

• memory module makers

  Listed as the least favored segment.

  Strengths

  Could still generate cyclical gains if terminal restocking resumes or price momentum continues.

  Weaknesses

  The report sees this segment as not being at the core of the strongest capital flows and bottlenecks.

  Comparison

  Least favored in the report relative to DRAM, legacy memory, and NAND.

  Risks

  Margin pressure, inventory risk, weak price pass-through, and demand volatility.

• Samsung Electronics

  One of the core Korean technology/memory-related names covered and disclosed in the report.

  Strengths

  Benefits from long-term AI memory demand and a potentially extended upward storage cycle.

  Weaknesses

  Short-term stock performance may be pressured by sector crowding, hyperscaler volatility, and slowing price momentum.

  Comparison

  Along with SK hynix, it is a core disclosed storage-related name for the report, and both are Overweight.

  Risks

  Product and storage-cycle concentration, Apple and Chinese smartphone competition, and semiconductor earnings concentration risk.

• SK hynix

  One of the core Korean storage-related names covered and disclosed in the report.

  Strengths

  Highly beneficiary of AI memory demand, DRAM demand strength, and hyperscale data center growth.

  Weaknesses

  Holding has become crowded after a strong run; near-term valuation and price momentum face resetting.

  Comparison

  Along with Samsung Electronics, both are Overweight disclosed names, but SK hynix has higher sensitivity to the DRAM cycle.

  Risks

  DDR5 competition causing excessive supply-side spending, elevated inventories at cloud and Chinese smartphone customers, and demand weakening.

Key data

• Report date2026-07-06The cover page time is July 6, 2026 07:34 PM GMT.
• 2027e earnings growthabove 35%-40%The report cites this as one of the main reasons for its continued long-term bullish stance on memory.
• Current priceSamsung Electronics 318,000.00 KRW; SK hynix 2,343,000.00 KRWReferenced stocks share prices disclosed on page 7 of the report.
• Industry viewAsia Pacific Industry View: AttractiveThe report discloses Attractive as the industry view for S. Korea Technology.
• Valuation assumptionsCost of equity 11.5%, risk-free rate 5%, equity risk premium 6.5%, beta 1.0, perpetual growth rate 3%From the valuation methodology and risk disclosure section.
• NAND price outlookTotal NAND Flash: 1Q26 up 85%-90%, 2Q26 up 55%-60%, 3Q26E new forecast up 10%-15%, 4Q26E new forecast up 0%-5%From the latest DRAM and NAND price forecast extract in the report table.

Impact & implications

For investment implications, the report does not interpret the near-term correction as the start of a new bear market, but more as a necessary reset after crowded positioning and a cyclical peak in the rate of change. Portfolio-wise, greater focus should be on DRAM and legacy memory, where AI capital flows are most direct and supply-demand bottlenecks are clearer, while remaining cautious on NAND, memory module makers, and chains more sensitive to low-cost tokens and open-source models.

Risks

• If the major AI capex participants truly have excess compute, the market may continue the narrative of AI overbuild.
• If hyperscalers cut capex after 2Q26 earnings or issue weaker-than-expected 3Q guidance, storage stock performance could be pressured.
• Token minimization, low-cost open-source LLMs, and Chinese open-source models could weaken spending expectations tied to higher-cost frontier models.
• DRAM year-over-year price, inventories, and breadth of earnings revisions are near peaks, so stocks may correct before fundamental changes are fully reflected.
• The storage segment is crowded, and higher volatility may force investors to reduce high net exposure.
• Terminal demand below expectations, DDR5 competition leading to excessive supply-side investment, and elevated inventories at cloud and Chinese smartphone customers.

What to watch

• Whether hyperscaler earnings and capex in 2Q26 are maintained or increased.
• Whether AI spenders’ 3Q26 revenue, margin, and token-demand guidance is raised or maintained.
• Spot and contract prices for DRAM and NAND, and year-over-year price change rates.
• Beyond earnings commentary from memory companies, the market’s real reaction to customer-side capex and inventories.
• Whether revision breadth remains at historical highs or starts to roll over.
• Whether crowded positioning continues to de-risk and whether capital rotates toward laggards such as MLCC and semicap.