Daily Brief

Daily perspectives and analysis from institutional research
← Back to daily briefs
Daily Brief

2026-09-01 Daily Quick Read | Heluo Investment Research

Summary

The current market is driven by two engines: AI infrastructure investment and macro structural transformation. AI computing demand is spilling over from standalone HBM to full-stack memory architectures, data center power equipment, and high-end PCB materials, while accelerating cloud API monetization validates the commercialization path for large models. At the macro level, the U.S. economy shows no broad overheating but faces localized capacity constraints; China's manufacturing PMI has improved marginally, while real estate policy shifting toward completed-home sales is reshaping the industry landscape. Humanoid robot forecasts have been significantly revised upward, and digital advertising, empowered by AI, is seeing both volume and price rise, but global trade frictions and geopolitical compliance costs are materially suppressing the profit margins of companies expanding overseas.

2026-09-0153 reports6 institutions
Published: Content updated:
01

AI Memory Architecture Evolution and Storage Pricing Cycle

3 Related reports

Key views

Bernstein constructed a complete AI memory architecture hierarchy from on-chip SRAM, HBM, system DRAM, and CXL to local SSDs, noting that the decoding phase of Transformer inference is memory-bound because KV cache grows linearly with tokens and concurrent users, and RAG databases and agent workflows further amplify demand for large-capacity storage and system DRAM.

Goldman Sachs data shows that in 8 month, DRAM spot prices remained strong, with DDR4 rising 7% month-on-month and trading at a 43% premium to contract prices; South Korea's 7-month DRAM exports surged 394% year-on-year, hitting a 2008-year high. It also significantly raised SK Hynix's 2027 HBM pricing growth forecast from 50% to 100%, projecting its 2027 HBM revenue to reach USD 630 hundred million.

New memory technologies face different commercialization challenges: HBF, led by SanDisk and SK hynix, requires a substantial leap in NAND performance; Samsung's zHBM faces doubts over heat dissipation and yield; NVIDIA's NVHBM could undermine memory suppliers' base-die differentiation and shift value to foundries such as TSMC; while PIM is constrained by ecosystem migration costs.

Traditional DRAM ASP gains are slowing. Goldman Sachs expects 3Q26E industry DRAM ASP to grow roughly 17% sequentially, but the second-order change is about negative 20 percentage points; affected by high inventory, mobile pricing growth in 3Q26 is expected to slow to 8-13%.

Current market environment

AI inference demand is shifting from compute bottlenecks to memory bandwidth and capacity bottlenecks. Extremely tight HBM supply and demand support forward pricing, while traditional DRAM, though tight in the spot market, already shows signs of slowing price momentum.

Future market changes

Widespread Adoption of NVHBM Architecture Reshapes Profit Distribution in the HBM Supply Chain

Long term

Triggers

  • NVIDIA fully adopts its self-developed base-die NVHBM solution in next-generation accelerators

Transmission channels

  • Memory controller functions moved to the GPU base die
  • Memory suppliers lose base-die differentiation capability
  • Manufacturing value concentrates in wafer foundries such as TSMC
  • HBM supplier margins under pressure

Indicators to watch

  • Release specification confirmation of NVIDIA's new architecture products
  • Changes in TSMC advanced packaging capacity allocation

Invalidation conditions

  • Memory suppliers successfully retain customized base-die design rights
  • NVHBM solution shelved due to yield or cost issues

Institutional disagreements

Sustainability of traditional DRAM ASP increases

Different views

  • Goldman Sachs believes spot premiums and server demand will support steady DRAM price increases, and suppliers prioritizing capacity allocation to HBM/server DRAM will create supply constraints.
  • TrendForce data shows that high mobile inventory has temporarily weakened procurement momentum; mobile DRAM pricing growth in 3Q26 will slow significantly to 8-13%, and further decline to 0-5% in 4Q26.

Opportunities and risks

Investing in HBM long-term agreement premiums

Consensus opportunity

SK Hynix's 2027 HBM pricing growth forecast was significantly raised to 100%; tight supply and demand combined with favorable long-term agreement terms support above-consensus returns.

Potential beneficiaries

  • SK Hynix
  • Equipment vendors in the HBM industry chain

Risks

  • Excessive diversion of traditional DRAM capacity to HBM leading to overall supply imbalance
  • AI capital expenditure falling short of expectations

Indicators to watch

  • SK Hynix quarterly HBM revenue and ASP data
  • Capital expenditure guidance from major cloud providers
Related reports(3)

This content is compiled from institutional research report views, is for research reference only, and does not constitute investment advice.

Zhejiang ICP No. 2022035445-5
Disclaimer: Market data, charts, indicators, research views, and other information provided on this website are intended solely for information display, research communication, and educational reference. They should not be regarded as personalized investment advice, securities recommendations, trading instructions, solicitations, or guarantees of return. While we strive to improve the reliability of our data and content, such information may still be subject to delays, errors, incompleteness, or untimely updates due to source differences, methodological limitations, system processing, or market volatility. Users should exercise independent judgment based on their own circumstances and bear all risks and responsibilities arising from the use of this website.

Settings

Sign in to view recent logins