UBS raises second-half memory pricing assumptions and sees LTA negotiations as continued support for the upside in DRAM, NAND and HBM cycles.
AI summary card
UBS raises second-half memory pricing assumptions and sees LTA negotiations as continued support for the upside in DRAM, NAND and HBM cycles.
The report argues that the memory industry supply-demand gap remains in an unusually tight, historically rare state, and the recent pullback in memory stocks is more of a temporary adjustment after crowded trading than a fundamental reversal.
- UBS raised its 3Q26 DDR contract pricing assumption to +32% QoQ and 4Q26 to +18% QoQ.
- NAND 3Q26 pricing assumption was raised to +30% QoQ, while 4Q26 remains at +12% QoQ.
- UBS expects DRAM to remain undersupplied at least through 2Q28, with 2027 demand growth of +36.2% versus supply growth of only +19.3%.
- HBM demand forecast was slightly raised to 33.1bn Gb in 2026 and 58.7bn Gb in 2027.
- Memory shares have fallen by about 17% on average from June highs, but UBS still sees strong fundamentals and segment 2027E free cash flow near US$1.2tn.
Report interpretation
Overview
This is a UBS global semiconductor monthly industry report focused on memory pricing in the second half of 2026, LTA long-term agreement negotiations, HBM demand forecasts, and investment views after the recent pullback in memory stocks. The report argues that DDR and NAND price upside is higher than previously expected, DRAM supply-demand tightness may persist at least through 2Q28, and the NAND upcycle may continue at least through 4Q27.
Core views
UBS's core view is that the current memory-cycle upside is driven by supply constraints, expanding AI server and HBM demand, and DDR5 long-term agreement demand and price lockups. Even though several memory names have pulled back about 17% from June peaks, the report still sees this as reflecting crowded positioning and near-term sentiment rather than a deterioration in sector fundamentals. News about Meta potentially selling compute capacity is seen as not affecting HBM procurement, as it is more likely asset monetization of legacy assets and positioning ahead of the rollout of next-generation Blackwell/Rubin and MTIA capacity.
Analysis framework
The report primarily uses industry research, supply-demand modeling, pricing-assumption revisions, LTA negotiation tracking, bottom-up HBM end-demand estimation, and stock-rating and target-price frameworks. Supply-demand analysis covers DRAM, NAND, and HBM; the demand side is further decomposed into Nvidia AI GPU, Google TPU, AMD, AWS, and hyperscale cloud provider procurement.
Methodology notes
Supply-demand gap and sufficiency ratio
By comparing bit demand growth, bit supply growth, and inventory drawdown assumptions, the report assesses the degree of undersupply in the industry and the persistence of the pricing cycle.
DDR and NAND sequential contract ASP assumptions
Based on industry research and progress in LTA negotiations, UBS raised the 3Q26 and 4Q26 DDR contract price assumptions and raised the 3Q26 NAND price assumption.
HBM demand tied to AI GPU/TPU purchases
Based on accelerator demand from Nvidia AI GPUs, Google TPUs, AMD, AWS, and other customers, UBS derives HBM Gb demand forecasts.
Target-price approaches for memory makers
Samsung, SK Hynix, and Nanya Tech use a 12-month forward P/BV or PBR framework, Micron uses SOTP valuation, and Kioxia uses a PBR-based target price.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- Samsung ElectronicsCore beneficiary and Key Call Buy name
- Strengths
- Upward cycle in DDR5 and HBM, progress in LTA agreements, and a target price of Won550k.
- Weaknesses
- High consumer electronics exposure, potential earnings pressure from won appreciation, and governance transparency risks at the group level.
- Comparison
- Compared with other memory peers, Samsung spans DRAM, NAND, HBM, and end-market products, combining cyclicality leverage with business diversification.
- Risks
- Memory cycle pullbacks, slower smartphone demand, capex volatility, and company/government and market risks in Korea.
- SK HynixPrimary beneficiary of the HBM and DRAM upward cycles
- Strengths
- Buy rating, target price raised from Won3.00m to Won3.20m, benefiting from HBM and DDR5 demand.
- Weaknesses
- Still exposed to memory commodity cycles and capital-intensive industry dynamics.
- Comparison
- Compared with peers with more diversified NAND exposure, SK Hynix has higher leverage to the HBM and DRAM cycles.
- Risks
- Volatility in smartphone, tablet, and enterprise spending, and cash-flow swings from rapid contract price moves.
- Micron Technology IncUS memory manufacturer; report retains Buy
- Strengths
- Target price US$1,625, benefiting from DRAM, NAND, and improving AI memory mix.
- Weaknesses
- High volatility in the memory cycle with earnings tied to global GDP and ASP changes.
- Comparison
- Compared with peers in Asia, Micron is US-listed and applies an SOTP valuation framework.
- Risks
- Macro demand deterioration, larger-than-expected ASP declines, slower-than-expected technology transitions, and timing shifts when supply increases.
- KioxiaBeneficiary of the NAND upcycle
- Strengths
- Buy rating, target price ¥132,000, directly benefiting from rising NAND prices.
- Weaknesses
- The business is more concentrated in NAND, making it more exposed to single-market cycle swings.
- Comparison
- Compared with firms with higher DRAM/HBM weighting, Kioxia is more sensitive to NAND cycle fluctuations.
- Risks
- High NAND cyclicality, price declines after supply-demand balance is restored, AI spending normalization, and uncertainty over sustained technology advantage.
- Nanya TechnologyTaiwan pure DRAM maker; report maintains Buy
- Strengths
- Target price NT$495, benefits from higher DRAM prices, and the table shows a Buy rating on 2026-06-08.
- Weaknesses
- Its product mix is more traditional DDR-oriented and more vulnerable when end-demand slows.
- Comparison
- Compared with Samsung, SK Hynix, and Micron, Nanya is more a pure DRAM and traditional-cycle play.
- Risks
- End-demand slowdown, downside cycle in traditional memory demand, and an overly legacy-centric product mix.
- META PLATFORMS INCHBM-demand-side customer, not a core rated name in this report
- Strengths
- AI compute investments remain an important source of memory demand.
- Weaknesses
- News about selling compute assets briefly pressured memory-stock sentiment.
- Comparison
- UBS views Meta-related coverage as more likely monetization of existing assets rather than a weakening of HBM procurement demand.
- Risks
- If AI capex or compute procurement cadence slows, upstream HBM demand expectations could be affected.
- ADVANCED MICRO DEVICES INCRelated AI accelerator ecosystem demand source
- Strengths
- The report raised 2027 AMD-related HBM procurement assumptions.
- Weaknesses
- Its impact is mainly in upstream HBM demand modeling rather than direct stock-level ratings in this report.
- Comparison
- Like Nvidia and Google TPU, AMD accelerator shipments are one input in the HBM demand model.
- Risks
- AI GPU competitive dynamics, customer procurement cadence, and advanced-packaging supply could affect HBM pull-through.
Key data
- 3Q26 DDR contract price assumption+32% QoQPreviously +17%; 2Q26 was +67% QoQ.
- 4Q26 DDR contract price assumption+18% QoQPreviously +12%.
- 3Q26 NAND price assumption+30% QoQPreviously +17%; 4Q26 remains at +12% QoQ.
- DRAM demand and supply growth2027E demand +36.2% YoY, supply +19.3% YoYUBS believes the gap is too large and unlikely to close until before 2Q28 at the earliest.
- HBM demand forecast2026E 33.1bn Gb; 2027E 58.7bn GbThese correspond to +90% YoY and +77% YoY, respectively.
- Nvidia AI GPU-linked HBM procurement assumption2026E 8.5 million units; 2027E 11.0 million unitsConsistent with UBS's recent revision to CoWoS estimates.
- Google TPU-equivalent units for HBM2026E 4.2 million units; 2027E 9.1 million unitsThe report increased its 2027E Google TPU-related assumptions.
- Memory industry revenue forecast2026E US$992bn; 2027E US$1,763bnReflects revenue expansion from higher prices and tighter supply-demand balance.
- Memory industry free cash flow2027E near US$1.2tnUBS believes this could support higher shareholder returns.
Impact & implications
If UBS's call is confirmed, earnings, cash flow, and shareholder return expectations for memory makers still have upside, particularly for leading peers with HBM, DDR5, and LTA pricing-lock capabilities. For downstream customers, the biggest pressure is memory costs and the affordability of AI capex; hyperscale cloud providers may need to continue raising funds in capital markets to fund continued investment.
Risks
- Affordability by customers is the primary risk to this upside cycle, particularly because hyperscale cloud providers must continue financing to support AI capex.
- The memory sector remains cyclical by nature, and DRAM and NAND prices could correct once supply-demand balance is restored.
- If macro deterioration damages demand, target prices for firms such as Micron face downside risk.
- Slower AI investment, lower-than-expected server demand, or changes in customer inventory burn could weaken HBM and DRAM demand.
- Technology transition, advanced packaging, HBM supply ramp, and persistence of NAND technological advantages all carry execution risks.
- Samsung, SK Hynix, and others also face FX, governance, domestic market, and capex volatility risks.
What to watch
- The progress of LTA signing around 3Q26 for Samsung and other Korean memory makers, especially progress toward DDR5 lockup targets of 50%-70%.
- Outcomes of SK Hynix negotiations with large hyperscale customers on DDR5 and NAND flash.
- Whether 3Q26 and 4Q26 DDR and NAND contract prices materialize to match UBS's raised assumptions.
- Whether the gap between 2027 DRAM demand growth of +36.2% and supply growth of +19.3% persists.
- Whether HBM demand assumptions tied to Nvidia AI GPUs, Google TPUs, AMD, and AWS continue to be raised.
- Changes in crowding/tightness after the pullback in memory shares from the June high, and whether improved free cash flow translates into higher shareholder returns.