China should support AI advancement with power, computing resources, premier says

Chinese Premier Li Qiang urged better coordination of power and computing resources to accelerate scaled, commercialised applications of artificial intelligence and called for an improved environment for AI firms and talent, plus expanded international technology exchanges. The comments signal Beijing-level support that could boost demand for data centers, cloud and chip providers and increase power infrastructure investment, but lack immediate fiscal or regulatory specifics, limiting near-term market-moving impact.

Analysis

Market structure: State-led coordination of power + compute is a direct demand shock for domestic AI cloud, data‑centre operators, and local fabs (BIDU/BABA/SMIC/1088.HK). Winners: AI-platform/cloud providers and utilities that sell firm power or captive coal; losers: foreign GPU suppliers if export controls tighten and ad/gaming incumbents whose monetization is slower. Increased compute intensity implies sustained upward pressure on electricity and thermal coal demand by mid-2026, shifting pricing power to energy suppliers and data‑centre capacity owners.

Risk assessment: Tail risks include stricter US export controls, Beijing backtracking on cross‑border tech exchanges, and grid bottlenecks causing brownouts or caps—each can wipe out revenue assumptions within 3–12 months. Immediate moves (days) will be sentiment-driven; over 6–18 months execution risk (permits, land, grid upgrades) dominates; 2–5 years sees meaningful domestic substitution of Western chips. Hidden dependencies: cooling/water, local government power quotas, and specialized server supply chains.

Trade implications: Tilt portfolios toward AI-native cloud names (BIDU/BABA), tactical longs in power/coal (China Shenhua 1088.HK) for 3–6 months, and selective fab exposure (SMIC) only after subsidy clarity. Use pair trades to isolate AI upside vs ad/gaming cyclicality (long BIDU, short TCEHY) and small convex option bets (NVDA calls) to capture supply repricing.

Contrarian angles: Consensus underestimates implementation drag—monetization of AI services historically takes 2–4 years; immediate rally in broad China tech could be overdone. Unintended consequences: rapid capex could trigger environmental pushback or reallocation away from consumer internet, compressing margins for incumbents; watch for policy detail that ties power allocation to local industrial priorities.

Key Decisions for Investors

• Consider establishing a 2–3% long position in Baidu ADR (BIDU) over 6–12 months, target +35% upside, set stop-loss at -15%; rationale: primary beneficiary of state compute & AI commercialisation support.
• Initiate a 1–2% tactical long in China Shenhua (1088.HK) for 3 months to capture higher power/thermal coal margins; take profit at +25% or cut loss if thermal coal futures drop >20% from entry.
• Implement a 2% long BIDU / 2% short Tencent ADR (TCEHY) pair trade over 6–12 months to express relative benefit to AI-native monetization versus ad/gaming exposure; unwind if the spread compresses to within 10% of 12‑month historical mean.
• Buy NVDA 3‑month 10% OTM calls sized to 0.5–1% of portfolio notional as convexity exposure to accelerated global AI GPU demand; if implied volatility rises >50% from entry, sell half to realize gains; maximum loss = premium.
• Conditional trade: Allocate 1–2% to SMIC (0981.HK) only if central government announces a semiconductor infrastructure subsidy package ≥ RMB 100bn within next 6 months; otherwise avoid due to export‑control tail risk.