Citi Names 7 China Data Centre Stocks as AI Buildout Accelerates

Citi has identified seven stocks as the clearest equity plays on China’s national data centre programme, splitting them across two categories with distinct risk and timing profiles. Two of those names, GDS Holdings and VNET Group, have already posted record bookings in Q1 2026, figures the bank says confirm that policy is translating into real commercial activity. The framework, reported via Investing.com, gives investors a structured entry point into a multi-year, state-driven infrastructure cycle that spans data centre operators, AI server manufacturers, optical component makers, and software services providers. What follows profiles all seven stocks Citi flagged, explains the two-category logic underpinning the recommendations, and identifies the specific data points and watch variables that matter for each position.

Why Citi split China’s data centre opportunity into two distinct categories

Citi’s framework rests on a distinction that changes how each stock should be evaluated. The bank separates beneficiaries into data centre operators and AI supply chain localisation beneficiaries, reflecting materially different demand timing and risk structures.

• Data centre operators benefit from contracted, multi-year revenues driven directly by new compute infrastructure demand. Revenue visibility is relatively high because hyperscaler and enterprise leases lock in capacity commitments for years.
• Supply chain localisation names depend on domestic technology mandates that ring-fence procurement from foreign competitors. The opportunity is broader but more execution-dependent, tied to production ramp-ups, procurement cycles, and software deployment timelines.

Understanding which category a name belongs to is the prerequisite for assessing its timeline and risk. Treating all seven as interchangeable infrastructure plays would be a misread of the framework.

Data centre operators: early wins for GDS and VNET

The operator thesis is grounded in numbers, not policy projections. Both GDS and VNET posted record booking quarters in Q1 2026, and Citi’s reading is that these figures confirm an existing demand acceleration rather than a one-off event driven by programme announcements.

The record bookings at GDS and VNET sit within a broader surge in global AI capital expenditure, with the four largest US hyperscalers collectively committing approximately $725 billion in 2026 alone, a spending trajectory that is pulling forward data centre capacity demand across multiple geographies simultaneously.

Q1 2026 Booking Comparison GDS Holdings: approximately 200 megawatts of new wholesale bookings (record quarter). VNET Group: in excess of 500 megawatts in new orders (record quarter). Combined, these bookings represent contracted future revenue that will compound over multiple years as capacity is built and brought online.

GDS Holdings

GDS recorded approximately 200 megawatts of new wholesale bookings in Q1 2026. At wholesale scale, these leases typically represent multi-year commitments from top-tier cloud tenants, creating durable contracted revenue.

Citi noted that GDS had already been capturing accelerating demand momentum before the programme’s formal rollout, meaning the national initiative acts as a force multiplier on an existing upcycle. The watch variable is GDS’s capacity mix: how new builds split between domestic AI workloads and broader enterprise use will influence long-term pricing power and capital expenditure requirements.

VNET Group

VNET secured in excess of 500 megawatts in new orders during the same period. The company operates a mixed retail and wholesale model across core urban markets.

As with GDS, VNET’s bookings were already inflecting higher before the programme launched. The key concern is different, however. VNET has historically operated with higher leverage than some peers, and as demand for new capacity accelerates, balance sheet fitness and financing costs become as material as the top-line booking figures themselves.

How to read a state-driven infrastructure programme as an equity catalyst

The China data centre programme provides a live case for a framework that applies well beyond this single theme. Whether evaluating state-backed programmes in energy, transport, or defence, the same analytical steps apply.

1. Map the value chain. Identify beneficiaries from physical infrastructure through compute hardware, connectivity, and software. Each layer carries different demand timing, margin profiles, and risk characteristics.
2. Distinguish near-term from longer-horizon beneficiaries. Operators with contracted demand sit at one end. Supply chain names tied to localisation and capacity ramp-ups represent a longer-horizon opportunity with more execution variables.
3. Anchor expectations in observable operating data. Booking trends and order backlogs provide more reliable signal than top-down spending headlines. The record Q1 2026 bookings at GDS and VNET are a concrete example of confirmatory data outweighing policy announcements as a conviction driver.
4. Assess balance sheet fitness for the growth cycle. Capital-intensive buildout phases require efficient access to funding. Leverage, credit ratings, and financing costs become material inputs during rapid expansion.

The programme’s stated five-year horizon is a planning guide, not a guaranteed backlog. Government mandates that create demand can also be delayed, redefined, or reprioritised.

Supply chain plays: the localisation mandate for five key names

China’s programme explicitly embeds domestic technology mandates designed to reduce reliance on foreign semiconductor and technology supply chains. Citi identifies five companies across servers, networking, and software as proxies for the localised AI ecosystem. Each occupies a different layer of the data centre stack.

China’s $295 billion AI buildout plan targets domestic suppliers for 80% of technology procurement, including AI chips, which is the policy foundation that transforms localisation mandates from regulatory guidance into a guaranteed addressable market for the five supply chain names Citi identifies.

IEIT Systems

• Investment thesis: Direct AI server proxy. IEIT’s server business is currently operating under supply constraints, indicating demand outpaces current capacity. The domestic chip mandate adds a policy-backed demand layer on top of existing commercial orders.
• Watch variable: Production scale-up execution, including capex guidance and the ability to ramp output without sacrificing margins.

The domestic chip mandate underpinning IEIT’s investment thesis is partly a consequence of Nvidia’s near-total exit from the Chinese AI accelerator market, with Huawei Ascend now holding an estimated 70-80% share and creating structural procurement demand for local server manufacturers built around non-US silicon.

Lenovo Group

• Investment thesis: The Infrastructure Solutions Group (ISG) offers diversified exposure to domestic server localisation without pure-play concentration. Data centre demand represents incremental upside alongside PC and mobile revenues.
• Watch variable: ISG segment revenue growth and margin trends, particularly any disclosures tying order growth explicitly to AI and data centre projects.

ZTE Corporation

• Investment thesis: AI server and data centre networking exposure with a structural procurement advantage via established relationships with state-owned enterprises and government-linked customers.
• Watch variable: Geopolitical and regulatory headline risk stemming from ZTE’s prior history with U.S. export restrictions. Escalating U.S.-China technology tensions could introduce valuation volatility even as domestic demand strengthens.

Accelink Technologies

• Investment thesis: Optical transceiver demand scales directly with data centre capacity and generates recurring upgrade demand beyond initial builds. Accelink is a leading domestic manufacturer in this segment.
• Watch variable: Market share in domestic optical segments and the impact of higher-speed standards (800G and beyond) on product mix and margins.

Chinasoft International

• Investment thesis: Software and services layer deeply integrated with Huawei’s cloud and AI platforms. Positioned for the localised software stack buildout that accompanies hardware deployment.
• Watch variable: Revenue concentration risk tied to Huawei. The company’s growth trajectory depends heavily on Huawei’s competitive position in domestic cloud and AI infrastructure.

Cross-cutting risks that apply to all seven names

No state-driven infrastructure thesis is without structural risks, and investors who enter these positions without accounting for them face foreseeable volatility.

• Policy execution risk: The same government mandate that creates demand can be delayed or redefined. The five-year programme horizon should be treated as a planning guide, not a committed backlog.
• Timing divergence between categories: GDS and VNET are capturing contracted revenue now. Supply chain names face a more gradual ramp tied to procurement cycles, production capacity scaling, and software deployment timelines.
• Geopolitical and regulatory risk: ZTE’s history with U.S. export restrictions is the clearest example of this risk materialising for a name on the list. Broader U.S.-China technology tensions remain a source of valuation volatility across the entire group.

U.S.-China technology tensions remain structurally durable regardless of bilateral trade progress because AI chip export controls are grounded in national security law with bipartisan Congressional backing, placing them outside the jurisdiction of trade negotiators and keeping ZTE’s regulatory exposure elevated even in optimistic diplomatic scenarios.

The stated five-year programme horizon is a planning guide, not a guaranteed backlog. Policy risk applies across both categories.

What Citi’s framework signals about the broader China AI infrastructure trade

The combination of record operator bookings in Q1 2026 (GDS at 200 megawatts, VNET at over 500 megawatts) and embedded domestic localisation mandates means the China data centre programme has moved from policy aspiration to observable commercial activity across multiple value chain layers.

Citi’s two-category structure is itself a signal of the programme’s maturity. The ability to segment the opportunity by timing and risk reflects enough on-the-ground data to make category-level distinctions, a development that was not available when the programme sat at the announcement stage.

The five supply chain names function collectively as a localised ecosystem proxy, benefiting from structural demand ring-fencing that limits foreign competition across servers, networking, and software. For investors, this framework offers both a specific watchlist and a transferable method for evaluating future state-backed infrastructure cycles.

For investors wanting to apply this value chain framework beyond China’s national programme, our deep-dive into AI supply chain investing maps where profit is genuinely concentrated across semiconductors, foundries, cloud operators, and software globally, including analysis of how legacy application software multiples are diverging sharply from AI-native valuations.

This article is based on Citi’s research as reported by Investing.com. It is for informational purposes only and should not be considered financial advice. Investors should conduct their own research and consult with financial professionals before making investment decisions.