China CIO Survey Supports Cloud Semiconductor and Edge AI Sentiment; Upside for Legacy Memory Prices Limited

Overview

Based on Morgan Stanley's 1H26 China CIO survey, this report assesses the impact of cloud computing, GenAI, physical AI investment, and legacy memory price increases on the Greater China semiconductor supply chain. Survey results are broadly positive: AI/ML/PA ranks as the top priority for incremental CIO spending, data center build-out ranks third, and public cloud deployment alongside edge AI-related spending continues to improve.

Core views

The core view is that demand for cloud and edge AI remains robust, benefiting cloud semiconductor and edge AI-related companies; however, rising legacy memory prices have begun to influence customer procurement behavior. While DDR4 demand remains solid, upside is constrained by customer budget tolerance and configuration optimization efforts. The report maintains Overweight views on Aspeed and Montage, citing sustained cloud spending, and maintains an Overweight view on Espressif due to its edge AI exposure.

Analysis framework

The report primarily utilizes CIO survey data to track IT spending priorities, intentions for GenAI public cloud deployment, the proportion of application workloads moving to the cloud, physical AI budget allocation, and procurement strategies in response to memory price hikes, mapping these findings to equity implications across the cloud semiconductor, edge AI, and legacy memory supply chains.

Methodology notes

• survey_analysisChina CIO Survey 1H26

  Assessing corporate IT demand trends through CIO spending intentions and deployment plans

  The survey covers CIO feedback on AI/ML/PA, digital transformation, data center build-out, public cloud, physical AI spending, and memory procurement strategies, used to infer shifts in semiconductor demand.

• Valuation methodsResidual income model

  Residual Income Model

  In the valuation methodology section, the report notes the use of the residual income model for select companies, listing key assumptions such as cost of equity, mid-term growth rate, terminal growth rate, and payout ratio.

Asset mapping & comparison

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

• Aspeed Technology (5274.TWO)

  Cloud semiconductor beneficiary

  Strengths

  The report believes sustained cloud spending supports its Overweight rating.

  Weaknesses

  Demand support could weaken if cloud capex or data center build-out slows.

  Comparison

  Compared to edge AI names, Aspeed offers greater exposure to cloud and data center semiconductors.

  Risks

  Tightening customer budgets, slowing cloud spending, or downward revisions to valuation assumptions.

• Montage Technology Co Ltd (688008.SS; 6809.HK)

  Cloud semiconductor and DRAM interface technology transition play

  Strengths

  Benefits from cloud capex growth, DRAM interface technology transitions, and localization of China's data center semiconductor ecosystem.

  Weaknesses

  Rising memory prices have made some customers more cautious in procurement, potentially limiting demand elasticity.

  Comparison

  Beyond pure legacy memory price elasticity, Montage also offers logic tied to interface technology transitions and localization.

  Risks

  Slower-than-expected DRAM interface technology transition, changes in the pace of US peer exit, or lower-than-expected cloud capex.

• Espressif Systems (688018.SS)

  Edge AI exposure play

  Strengths

  Increased physical AI spending by CIOs supports its edge AI demand thesis.

  Weaknesses

  The pace of edge AI adoption and realization of end-market demand remain to be seen.

  Comparison

  Compared to cloud semiconductor names, Espressif maps more directly to edge devices and physical AI investment.

  Risks

  Physical AI investment falling short of expectations, weakening end-demand, or intensified competition.

• DDR4/legacy memory

  Legacy memory pricing and demand theme

  Strengths

  The report views DDR4 demand as remaining resilient.

  Weaknesses

  CIOs are more cautious about price hikes, increasingly opting for optimization, deferral, or down-specification rather than direct budget increases.

  Comparison

  Compared to cloud semiconductors and edge AI, legacy memory is more constrained by customer budget tolerance.

  Risks

  Excessive price hikes leading to demand destruction, or shifts in supply and procurement timing causing price pullbacks.

Key data

• GenAI Public Cloud Deployment Plans54%Plans to implement GenAI via public cloud in the next 12 months, up from 44% in 2H25 and 28% in 1H25.
• Share of Application Workloads on Public Cloud31%CIOs estimate that 31% of application workloads currently run on public cloud, up from 30% in 2H25 and 24% in 1H24.
• Share of CIOs Allocating Budget to Physical AI63%Up from 57% in 2H25.
• Average Share of IT Spending on Physical AI5.7%Up from 3.8% in 2H25.
• Responding to Memory Price Hikes via Right-Sizing38%CIOs plan to adjust memory requirements to avoid over-provisioning.
• Responding to Memory Price Hikes by Directly Increasing Hardware Budgets20%Only a minority of CIOs chose to directly expand budgets.

Impact & implications

From an investment perspective, rising priorities for cloud spending and data center build-out continue to support cloud semiconductor demand, while expanding physical AI budgets underpin edge AI-related opportunities; however, in the face of memory price hikes, customers' preference for deferral, optimization, or down-specification suggests that further price increases for legacy memory may encounter demand elasticity constraints.

Risks

• Rising memory prices leading customers to defer purchases, down-specify, or purchase only critical capacity, potentially capping upside for legacy memory prices and volumes.
• Cloud capex, data center build-out, or GenAI public cloud deployment progressing slower than expected.
• Physical AI investment budgets failing to sustain growth, or edge AI demand materializing slower than anticipated.
• DRAM interface technology transition proceeding slower than expected.
• The report discloses that Morgan Stanley has investment banking or other business relationships with certain covered companies; investors should be aware of potential conflicts of interest.

What to watch

• Whether GenAI public cloud deployment plans in future CIO surveys continue to reach new highs.
• Whether the share of application workloads migrating to public cloud continues to rise above 31%.
• Whether the share of physical AI investment in IT budgets continues to climb from 5.7%.
• Whether CIOs continue to optimize configurations in response to memory price hikes or shift toward directly increasing hardware budgets.
• DDR4 and DDR5 price trends and customer procurement cadence.
• Subsequent rating changes, target price adjustments, and order trends for Aspeed, Montage, and Espressif.