CIO Survey: AI Tops Budget Priorities, Cloud Adoption Reaches New High

Overview

Morgan Stanley’s first-half 2026 China Chief Information Officer (CIO) Survey report indicates that corporate sentiment toward cloud computing and edge artificial intelligence (Edge AI) remains broadly positive. AI‑related technologies have emerged as the primary driver of increased IT spending, with a marked uptick in the willingness to deploy generative AI on public clouds. However, amid rising memory prices, most CIOs have adopted more prudent cost‑containment measures rather than simply expanding their budgets. Accordingly, the firm maintains an optimistic outlook on cloud‑semiconductor and edge‑AI‑related stocks, while taking a selective stance on the traditional memory sector.

Core views

AI and cloud computing have become absolute core priorities: Survey data show that AI, machine learning, and process automation (AI/ML/PA) top CIOs’ spending‑increase priorities for 2026, with a 37% share, while data‑center infrastructure ranks third at 8%. In terms of deployment strategies, the public cloud’s position has further solidified: 54% of surveyed CIOs plan to adopt generative AI via the public cloud within the next 12 months, up sharply from 44% in the second half of 2025 and 28% in the first half—marking a new high since the survey began. Currently, CIOs estimate that roughly 31% of application workloads run on the public cloud, a slight uptick from earlier levels. Rising share of physical AI investments: As AI adoption deepens, hardware spending is also expanding. At present, 63% of CIOs allocate their IT budgets to physical AI investments—such as servers and accelerator cards—up from 57% in the second half of 2025. On average, the proportion of IT spending devoted to physical AI is expected to rise from 3.8% to 5.7%, reflecting a shift from software algorithms toward underlying compute‑intensive infrastructure. Divergent responses to memory price hikes: Faced with recent memory price increases, CIOs have not uniformly opted to absorb the costs; instead, they have adopted a range of strategic measures. Only 20% of CIOs say they will directly increase their hardware budgets, while another 10% believe prices have not yet affected their purchasing behavior. The vast majority have chosen more nuanced approaches: 38% plan to address the issue by right-sizing memory requirements to avoid over‑provisioning; 17% are limiting purchases to only critical‑workload capacity and deferring non‑essential upgrades; 13% intend to delay procurement pending supplier concessions; 10% rely on software optimizations to improve memory efficiency; and the remainder are renegotiating contracts or opting for lower‑spec products. This cautious stance underscores that current memory pricing has become prohibitively expensive for some customers, thereby constraining upside potential for traditional memory technologies such as DDR4. Stock‑specific insights: Institutional investors maintain an “overweight” rating on Aspeed Technology (5274.TWO) and Montage Technology (6809.HK/688008.SS), citing the sustained growth in cloud‑related spending and the ongoing migration of DRAM interface technologies as key supporting factors. Similarly, Espressif Systems (688018.SS) retains an “overweight” rating, driven by its exposure to edge‑AI applications, which stands to benefit from increased CIO investment in physical AI infrastructure. For the traditional memory sector, while demand for DDR4 remains robust, institutional analysts note that CIOs’ cautious approach limits the upside from price increases, prompting a selective stance.

Analysis framework

This report employs a bottom-up, micro-level research approach, leveraging the AlphaWise platform to conduct a survey of chief information officers (CIOs) at Chinese enterprises, thereby gathering first-hand data on IT spending intentions. The analytical framework follows these key lines of reasoning: 1. **Demand-Side Validation**: By analyzing CIOs’ budgetary prioritization across various technology domains, we assess whether AI and cloud computing currently represent the most reliable growth drivers. 2. **Deployment Path Tracking**: We monitor shifts in the ratio between public cloud adoption and private/local deployments to gauge the extent to which cloud service providers and upstream chip manufacturers stand to benefit. 3. **Cost Sensitivity Analysis**: Focusing specifically on the impact of rising memory prices, we examine CIOs’ response patterns—whether they absorb cost increases or seek alternative technical solutions—to infer the pricing power of storage‑chip manufacturers and their sales outlook. 4. **Segmentation and Stock-Level Implications**: We map our macro‑level findings to specific semiconductor sub‑segments—cloud‑centric chips, edge‑computing MCUs, and storage‑interface chips—and, in conjunction with each company’s fundamental metrics, provide rating recommendations.

Methodology notes

• Industry/ Sector Analysis FrameworkSupply-and-Demand Framework

  By analyzing the purchasing intentions and budget allocations of downstream customers (CIOs), we can infer the demand outlook for the upstream semiconductor industry.

  Research reports do not directly forecast sales volumes; instead, they infer the true demand intensity and price elasticity across semiconductor sub-sectors—such as AI chips and memory—by probing “who is spending,” “where spending is occurring,” and “how companies are responding to price hikes.”

• Valuation MethodologyRIM Residual Income Model

  We employ the Residual Income Model to value Aspeed, Montage, and Espressif.

  This is a valuation approach that relies on both book value and projected excess returns. The research report meticulously outlines key assumptions, including the cost of equity (CoE), the mid-term growth rate, the terminal growth rate, and the dividend payout ratio, making it well-suited for assessing technology‑manufacturing companies with stable growth prospects and clear profit‑generating business models.

Asset mapping & comparison

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

• Aspeed Technology (5274.TWO)

  Benefiting from continued spending on cloud data center construction

  Strengths

  It holds a dominant position in the server remote management chip sector.

  Risks

  Weak cloud demand, slower-than-expected specification upgrades, and intensifying competition.

• Montage Technology (6809.HK / 688008.SS)

  Benefiting from the growth in cloud capital expenditures and the migration of DRAM interface technologies

  Strengths

  Strong position in the localization of data-center semiconductors in China

  Risks

  Cloud demand fell short of expectations, the migration of DRAM interface technologies has slowed, and new product launches have been delayed.

• Espressif Systems (688018.SS)

  Benefiting from opportunities in edge AI investments and increased spending on physical AI.

  Strengths

  It is competitive in the Wi‑Fi MCU sector and has entered edge AI applications.

  Risks

  The localization of MCUs is progressing more slowly than expected, new customer acquisition has fallen short of projections, and intensifying competition is eroding profit margins.

Key data

• AI/ML Spending Prioritization RankingsNo. 137% of CIOs have identified it as the top priority for increased spending in 2026.
• Public Cloud GenAI Deployment Intentions54%The plan is slated for implementation over the next 12 months, representing a significant increase from 44% in 2H25.
• Proportion of Physical AI Investment5.7%The average IT spending ratio has increased from 3.8% in 2H25.
• Strategies for Addressing Rising Memory Prices: Adjusting Scale38%The most mainstream strategy is to avoid over-allocation by right-sizing.
• Strategies for coping with rising memory prices: simply increase the budget.20%Minority shareholders indicate a higher price sensitivity.

Impact & implications

The survey findings reinforce the rationale behind “AI‑driven cloud capital expenditure,” benefiting semiconductor companies tied to cloud infrastructure—such as those producing server management chips and memory‑interface chips. Meanwhile, rising investment in edge AI is injecting fresh growth momentum into MCU and other edge‑computing chip manufacturers. However, the rapid surge in memory prices could dampen the release of certain non‑essential demand, leading to a weaker-than-expected recovery in the traditional storage market. Investors should remain vigilant about potential sales headwinds faced by memory vendors due to elevated pricing.

Risks

• Cloud demand has softened.
• Technical specification migration is progressing more slowly than expected.
• Intensified market competition
• China’s policy is tightening further.
• U.S. peers are exiting at a slower pace than expected (with respect to Montage).
• New product launch delayed

What to watch

• Progress in the Real-World Deployment of Public Cloud GenAI
• Trends in memory prices and changes in CIOs’ subsequent procurement behavior
• Quarter-over-quarter change in edge AI hardware shipments