Korean and Japanese memory makers enter the 1Q26 earnings season, with market focus shifting to AI demand resilience and supply constraints

Overview

This report reviews the key discussion points for the memory industry as Samsung Electronics, SK hynix, and Kioxia are about to release their 1Q26 results. JPMorgan believes that strong 1Q26 earnings are already largely reflected in market expectations, and that what truly needs to be tested is the sustainability of the memory upcycle after 2Q26, especially AI computing demand, customer price acceptance, long-term supply agreements, the speed at which capex converts into bit supply, and HBM product roadmaps and execution capability.

Core views

The core view is that the memory industry remains in a tight supply-demand environment and an AI-driven upcycle, but investors will shift from simply focusing on earnings revisions to testing the durability of profits. Whether customers can bear higher memory prices, whether AI model parameter expansion and larger KV cache/context windows continue to drive HBM demand, and whether long-term supply agreements can improve demand visibility will determine sector sentiment. The report also argues that capex headlines should not be simply interpreted as rapid supply release, because the complexity of advanced memory manufacturing, wafer consumption by HBM, equipment lead times, and infrastructure investment will all delay supply expansion.

Analysis framework

The report uses a combination of earnings preview and industry issue breakdown. It first judges that the market already expects strong 1Q26 results, then lists the questions investors need to ask across six dimensions: pricing, demand, technology, supply, HBM, and company events. The analysis focuses not on single-company valuation, but on judging the sustainability of the memory cycle and stock-price catalysts for the sector through industry developments at Samsung Electronics, SK hynix, Kioxia, Micron, and other global memory makers.

Methodology notes

• Industry cycle analysisTight supply-demand and upcycle framework

  A higher-for-longer memory upcycle

  The report combines AI demand, HBM supply bottlenecks, long-term supply agreements, and capex conversion efficiency to assess whether memory price increases and earnings improvement can continue beyond 2027E.

• Technology-driven demand analysismemory wall

  The performance gap between compute processors and memory speeds

  The report argues that token generation in AI computing, KV cache offloading, larger context windows, and model parameter growth will continue to reinforce demand for high-bandwidth and high-capacity memory.

• Contract and visibility analysisLong-term supply agreement LTA

  Enhancing visibility on demand, pricing, and earnings through long-term contracts

  Investors focus on LTA duration, volume-price structure, target profitability, and prepayment terms; the report places more importance on the process of suppliers establishing medium- to long-term demand visibility with key customers.

Asset mapping & comparison

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

• Samsung Electronics

  A major Korean memory supplier and one of the report's 1Q26 earnings names in focus

  Strengths

  Benefits from rising DRAM/NAND prices, AI memory demand, and tight industry supply; if strike impact is limited, fundamental pressure should be manageable.

  Weaknesses

  Investors are watching labor negotiations, incentive costs, and the impact of strikes on module/delivery links.

  Comparison

  Compared with SK hynix and Kioxia, Samsung's company-specific catalysts are more concentrated on labor costs, production impact, and execution recovery.

  Risks

  Strike impact exceeds expectations, OPEX rises, HBM execution lags peers, or customer price acceptance declines.

• SK hynix

  A major Korean DRAM/HBM supplier, with the report focusing on its earnings and ADR listing progress

  Strengths

  HBM demand expansion and AI server memory demand are favorable for the company, and share-price performance has been strong.

  Weaknesses

  Market expectations are high, so it needs to continue proving its HBM roadmap, pricing, and supply execution.

  Comparison

  Compared with Samsung, SK hynix investor focus is more skewed toward HBM execution and ADR listing updates.

  Risks

  HBM content growth or pricing comes in below expectations, wafer allocation pressure, or volatility in AI customer demand.

• Kioxia

  A Japanese NAND maker, with the report focusing on its earnings and early-June mid-term strategy briefing

  Strengths

  NAND spending expectations are relatively conservative, and if the gap between eSSD bit growth and CSP customer expectations narrows, there could be positive surprise.

  Weaknesses

  It needs to prove to the market its medium- to long-term demand outlook, LTA strategy, and shareholder return path.

  Comparison

  Compared with the DRAM/HBM core theme, Kioxia has more direct exposure to recovery in NAND and eSSD demand expectations.

  Risks

  NAND demand recovery falls short of expectations, the strategy briefing lacks clear catalysts, or the sustainability of price increases is insufficient.

• Memory sector

  The core industry asset basket in the report, including DRAM, NAND, HBM, and related supply chains

  Strengths

  Expansion of AI models, growth in KV cache and context windows, broader ASIC demand, and constrained supply expansion together support pricing and earnings.

  Weaknesses

  Valuation still reflects market skepticism about the sustainability of memory spending in 2027E.

  Comparison

  Figure 1 shows that multiple memory makers significantly outperformed the Philadelphia Semiconductor Index over the past year.

  Risks

  Hardware SRAM optimization or software compression reduces memory consumption, customers cut hardware spending, capex converts too quickly into supply, or LTA terms are unfavorable.

Key data

• Main earnings datesSKH Apr 23 9am KST; SEC Apr 30 10am KST; Kioxia May 15 4pm JSTThe earnings release dates for major Korean and Japanese memory makers listed in the report.
• 2027E total memory industry capex forecastUS$144bnJPMorgan says this forecast is up 74% from US$83bn in September 2025.
• Revision to 2027E DRAM/NAND bit supply assumptionsBoth revised up by 12-14%Despite a large upward revision to the capex forecast, the increase in bit supply assumptions is relatively limited, reflecting extended supply constraints.
• Market speculation on 2Q memory price increases30-50% Q/QThe report says that after strong pricing in 1Q26, the market is speculating about further price increases in 2Q.
• Price acceptance among some consumer electronics customersAgreed to proposed price hikes of 50% or moreThe report says price hikes may be more evident in consumer applications such as PCs and mobile than in servers.
• Estimated NVDA-level SOCAMM bit demand24-38bn Gb in 26E-27EJPMorgan's observation metric for SOCAMM2-related TAM and pricing.
• Global memory stock-price performance illustrationKioxia 1Yr +1715%; SNDK 1Yr +2988%; SK hynix 1Yr +606%; Micron 1Yr +578%; SEC 1Yr +299%; SOX 1Yr +158%Visual recognition data from Figure 1, showing that memory-related stocks significantly outperformed the SOX.

Impact & implications

If key customers continue to accept price increases and suppliers confirm AI memory demand, HBM content growth, and LTA progress on earnings calls, the valuation discount on memory stocks may narrow, and the market will have greater confidence in the sustainability of 2027E earnings. Conversely, if customer price resistance emerges, HBM content growth slows, LTA terms remain unclear, or capex is interpreted as releasing supply too quickly, sector sentiment may be pressured.

Risks

• Customers may be unable to continue accepting current or further higher memory prices, especially against the backdrop of scrutiny on major CSP free cash flow.
• Hardware SRAM optimization, software compression technologies, or system architecture changes may reduce memory consumption per unit of AI computing.
• If accelerated capex is interpreted by the market as future oversupply, it may pressure sector valuations.
• If LTA terms lack clarity on pricing, volume, profitability, and prepayments, they may fail to improve demand visibility.
• If HBM content growth, pricing, or yield execution disappoints, the AI memory thesis will be weakened.
• Samsung labor disputes may affect cost modeling, module/delivery processes, or investor confidence.

What to watch

• How SK hynix, Samsung Electronics, and Kioxia describe 2Q pricing, customer feedback, and order visibility in their 1Q26 earnings calls.
• The degree of acceptance by key customers of price increases in DRAM, NAND, HBM, and consumer-end memory.
• Whether suppliers reaffirm the memory wall thesis, and whether they are seeing signs of slowing AI memory content growth.
• LTA duration, volume-price structure, target profitability, prepayment arrangements, and disclosure cadence.
• How 2027E-2028E capex translates into DRAM/NAND bit supply, and the impact of HBM die penalty and equipment lead times.
• HBM4E content configuration, HBM pricing, and wafer allocation strategy between HBM and traditional DRAM.
• How Kioxia's early-June mid-term strategy briefing discusses demand, LTA, and shareholder returns.