Quick Summary
Covering the latest research from top Wall Street investment banks

China's RMB 2 trillion AI buildout plan is expected to significantly boost the domestic AI infrastructure ecosystem

Institution
Citi Research
Date
2026-06-09
Authors
Kyna Wong, Yiming Li, CFA, Karen Huang, Kevin Chen
Company
-
Ticker
-
Industry
Communications infrastructure, AI data centers, semiconductor localization
Rating
-
BullishLow confidenceThe report describes China's planned roughly RMB 2 trillion AI data center buildout as a major catalyst for the domestic AI infrastructure ecosystem, believing that IDC, AI servers, optical communications, domestic computing power services, and state-owned telecom operators will all benefit.
AuthorsKyna Wong, Yiming Li, CFA, Karen Huang, Kevin Chen
Asset classesEquity
Business segmentsNational data center network、AIDC capacity、Carrier-neutral IDC、State-owned telecom operators、Domestic AI chips、AI servers、Optical modules、Computing power services
Research firm divisions/subsidiariesCitigroup(Other)、Citigroup Global Markets Asia Limited(Other)

AI summary card

China's RMB 2 trillion AI buildout plan is expected to significantly boost the domestic AI infrastructure ecosystem

Citi believes that over the next five years, roughly RMB 2 trillion of investment in a national data center network, with no less than 80% localization in key technologies, will drive demand for AIDC capacity, IDC orders, domestic AI servers, optical modules, and computing power services.

Disclosed company ratings/prices: GDS.O at 1/US$34.69, VNET.O at 1/US$8.87, 0992.HK at 1/HK$25.38; 0354.HK, 000977.SZ, and 0763.HK at 2; 002281.SZ at 3. Under Citi's definitions, 1=Buy, 2=Neutral, 3=Sell; this report did not disclose target prices.
Industry researchAI infrastructureData centersLocalizationAI serversOptical modulesState-owned operators
  • According to a June 9 BBG report, the Chinese government plans to invest roughly RMB 2 trillion over the next five years to build a national data center network.
  • Based on an all-in cost assumption of about RMB 200 million/MW, the plan implies about 10GW of incremental AIDC capacity, or about 2GW/year.
  • Relative to China's total IDC installed base of about 20GW in 2026E, the plan is equivalent to increasing AI-grade capacity by about 50% from the current base.
  • GDS recorded about 200MW of wholesale orders in 1Q26, while VNET signed over 500MW of new orders in 1Q26; Citi believes demand momentum for carrier-neutral IDC has already begun to emerge.
  • The requirement that at least 80% of key technologies use domestic suppliers will benefit domestic supply chain segments such as Huawei's AI chip ecosystem, IEIT, Lenovo, ZTE, Accelink, and Chinasoft.

Report interpretation

Overview

This report focuses on China's communications infrastructure and AI data center industry chain. The core backdrop is a BBG report stating that the Chinese government plans to invest roughly RMB 2 trillion over the next five years to build a national data center network, led by the NDRC, mainly advanced by large state-owned telecom operators, and requiring that at least 80% of key technologies rely on domestic suppliers. Citi believes this plan will become a structural catalyst for China's digital infrastructure.

Core views

Citi's core views are: first, the RMB 2 trillion investment scale is sufficient to materially expand AI-grade data center capacity; second, even if the project is mainly advanced by state-owned telecom operators, carrier-neutral IDC providers will still benefit from accelerating demand for high-quality AI-ready capacity; third, the localization ratio requirement will promote the domestic ecosystem for AI chips, AI servers, optical modules, and computing power services; fourth, state-owned telecom operators will further transform from traditional connectivity service providers into foundational operators of China's AI economy.

Analysis framework

Starting from the scale of policy investment, the report uses an all-in cost assumption per MW to estimate incremental AIDC capacity, and then combines 1Q order momentum from GDS and VNET with the impact of localization ratio requirements on AI server, optical communication, and computing power service companies to map industry chain beneficiaries.

Methodology notes

  • Capacity estimationPer-MW all-in cost capacity estimation method

    RMB 2 trillion of investment divided by an all-in cost of about RMB 200 million/MW implies about 10GW of incremental AIDC capacity.

    This method incorporates both data center construction and AI server costs into a single unit cost assumption to estimate the scale of incremental AI data center capacity over the next five years.

  • Supply chain mappingLocalization beneficiary chain analysis

    At least 80% of key technologies rely on domestic suppliers.

    The report maps the policy localization requirement to segments including AI chips, AI servers, optical modules, computing power services, and state-owned telecom operators to identify potential beneficiary assets.

  • Order momentum validationIDC wholesale order observation

    Use GDS and VNET's 1Q26 order performance to validate demand for AI-ready data centers.

    GDS's roughly 200MW wholesale orders and VNET's more than 500MW of new orders are used to show that carrier-neutral IDC demand has already demonstrated strong momentum.

Asset mapping & comparison

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

  • GDS Holdings (GDS.O)
    Carrier-neutral IDC operator, benefiting from accelerating demand for AI-ready data center capacity.
    Strengths
    About 200MW of wholesale orders in 1Q26; the report believes it has already captured demand momentum.
    Weaknesses
    The plan is mainly advanced by state-owned telecom operators, and GDS's direct project share and cooperation model still need validation.
    Comparison
    Like VNET, it is a beneficiary in carrier-neutral IDC; disclosed rating is 1.
    Risks
    Policy implementation pace, customer delivery, capital expenditure pressure, rack utilization, and price competition.
  • VNET Group (VNET.O)
    Carrier-neutral IDC operator, benefiting from national data center network construction and demand for AI-ready capacity.
    Strengths
    More than 500MW of new signed orders in 1Q26, showing strong wholesale demand.
    Weaknesses
    Conversion of incremental orders into revenue and profit still depends on delivery, customer rack deployment, and financing capability.
    Comparison
    Like GDS, it represents IDC order momentum; disclosed rating is 1.
    Risks
    Construction cycle, funding costs, utilization ramp-up, customer concentration, and industry competition.
  • IEIT SYSTEMS (000977.SZ)
    Proxy for the domestic AI server ecosystem, benefiting from domestic AI chip content requirements and growing server demand.
    Strengths
    The report lists it as one of the key beneficiaries in the localized AI server ecosystem.
    Weaknesses
    The server business may face supply constraints, and profitability depends on chip supply, configuration, and pricing.
    Comparison
    Along with Lenovo and ZTE, it represents the domestic AI server theme; disclosed rating is 2.
    Risks
    AI chip supply, delivery cycle, gross margin volatility, and customer procurement pace.
  • Lenovo Group (0992.HK)
    AI server supplier, benefiting from data center construction and demand for localized AI servers.
    Strengths
    The report lists it as one of the proxies for the localized AI server ecosystem, and its disclosed rating is 1.
    Weaknesses
    The report did not disclose company-level financial weaknesses; the main uncertainties lie in the supply chain and order conversion.
    Comparison
    Compared with IEIT and ZTE, Lenovo is also a beneficiary of AI servers and carries a higher rating.
    Risks
    Server competition, supply of key components, gross margin, and AI project delivery pace.
  • ZTE (0763.HK)
    AI server and communications infrastructure supplier, benefiting from domestic AI infrastructure construction.
    Strengths
    The report lists it as one of the key beneficiaries of growing AI server demand.
    Weaknesses
    The report did not disclose specific weaknesses; the business is highly affected by telecom operator procurement cycles.
    Comparison
    Together with IEIT and Lenovo, it forms a domestic AI server beneficiary group; disclosed rating is 2.
    Risks
    Telecom operator capital expenditure volatility, tender pricing, supply chain, and policy execution pace.
  • Accelink Technologies (002281.SZ)
    Domestic optical module and optical communications supplier, benefiting from growth in optical connectivity demand within and between data centers.
    Strengths
    The report says that as a leading domestic optical transceiver company, it will benefit from a surge in optical connectivity demand.
    Weaknesses
    The disclosed rating is 3, indicating Citi is less positive on its stock return outlook than on some other beneficiary names.
    Comparison
    Unlike AI server companies, its benefits are more concentrated in optical interconnect and optical module segments.
    Risks
    Optical module price competition, technology iteration, customer qualification, demand realization, and a relatively cautious rating stance.
  • Chinasoft International (0354.HK)
    Huawei ITS and ISV partner, benefiting from domestic computing power services and token-based AI service demand.
    Strengths
    The report believes the plan will create a larger addressable market for its domestic computing power and AI services.
    Weaknesses
    Commercial realization depends on the depth of Huawei ecosystem cooperation, customer adoption, and conversion into service revenue.
    Comparison
    Compared with hardware suppliers, it benefits more from computing power services and the software ecosystem; disclosed rating is 2.
    Risks
    Service demand realization, project gross margin, ecosystem dependence, customer budgets, and the competitive landscape.
  • State-owned telecom operators
    Main operators of the national interconnected data center network, and transitioning into foundational operators of China's AI economy.
    Strengths
    They possess network resources, policy execution capability, and large-scale infrastructure operating experience.
    Weaknesses
    The report does not specify particular operators or project allocation, and earnings estimates remain unclear.
    Comparison
    Compared with carrier-neutral IDC providers, state-owned operators are more likely to undertake the primary construction and operating roles.
    Risks
    Return on capital expenditure, policy execution efficiency, procurement cycles, power resources, and business model profitability.

Key data

  • Policy investment scaleAbout RMB 2 trillion, approximately US$295 billionAccording to a June 9 BBG report, the plan is intended for building a national data center network over the next five years.
  • Project leadership and operationLed by the NDRC and mainly operated by large state-owned telecom operatorsThe report believes this will consolidate the role of state-owned telecom operators as foundational operators of China's AI economy.
  • Domestic key technology ratio≥80%Key technologies include AI chips from domestic suppliers such as Huawei, and the report believes this requirement will accelerate domestic semiconductor growth.
  • All-in unit capacity cost assumptionAbout RMB 200 million/MWIncluding data center construction and AI servers.
  • Implied incremental AIDC capacityAbout 10GW/five years, about 2GW/yearDerived from the RMB 2 trillion investment amount and the per-MW cost assumption.
  • China IDC installed baseAbout 20GW (2026E)The report states that the plan is equivalent to increasing AI-grade capacity by about 50% from the current base.
  • GDS 1Q26 wholesale ordersAbout 200MWThe report says its record wholesale orders reflect demand momentum for AI-ready IDC.
  • VNET 1Q26 new signed orders500MW+The report says its new order momentum supports the view that carrier-neutral IDC will benefit.

Impact & implications

If implemented, the plan will simultaneously change demand scale and supply chain structure: on the demand side, AI-grade data center capacity construction could accelerate materially, driving IDC, servers, optical interconnect, and computing power services; on the supply side, domestic AI chips and related ecosystems will receive stronger policy support due to the 80% localization requirement; on the operations side, the role of state-owned telecom operators in the national interconnected data center network will rise. For investment, opportunities are more concentrated in industry chain companies with orders, delivery capability, domestic ecosystem positioning, and computing power service capability.

Risks

  • The policy plan reported by BBG still requires validation through formal details, budget allocation, and execution pace.
  • The RMB 2 trillion investment translated using about RMB 200 million/MW may produce different capacity conclusions if unit costs, server configurations, or power constraints change.
  • The ≥80% localization requirement for key technologies may be constrained by AI chip capacity, performance, supply chain certification, and delivery cycles.
  • Since the plan is led by state-owned telecom operators, the actual order conversion and profitability of carrier-neutral IDC providers still need to be tracked.
  • AI server, optical module, and computing power service companies may face risks from tight supply, price competition, gross margin volatility, and customer concentration.
  • Regulation, export controls, sanctions, and cross-border investment restrictions may affect related securities and supply chains.

What to watch

  • Whether the NDRC and government departments publish a formal investment plan, funding sources, and construction timetable.
  • Tendering, capital expenditure, and procurement pace of state-owned telecom operators in the national interconnected data center network.
  • Supply, performance, and server adaptation progress of domestic AI chips, especially within the Huawei ecosystem.
  • Follow-up wholesale orders, delivered MW, utilization, and capital expenditure guidance from GDS and VNET.
  • AI server orders and delivery capability of IEIT, Lenovo, and ZTE.
  • Demand for Accelink optical modules, pricing trends, and volume growth in inter-data-center and intra-data-center connectivity.
  • Commercialization progress of Chinasoft's computing power services and token-based AI services.
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