China has too much green energy. Experts think AI can fix that

China is projected to account for 60% of global installed renewable capacity through 2030, and solar surpassed coal as the country's largest source of installed generating capacity in July. Renewable output effectively met all of China’s electricity-demand growth last year, but excess clean-energy infrastructure is creating overcapacity and increasing interest in AI-driven grid forecasting, storage optimization, and automated maintenance. Across Southeast Asia and India, grid fragmentation, continued coal dependence, and energy-security concerns remain constraints, although industry participants expect the regional clean-energy transition to continue.

Analysis

The key investable implication is not incremental renewable installation, but the widening spread between low marginal-cost generation and the system assets needed to monetize it. Persistent curtailment and depressed merchant power pricing would keep pressure on upstream solar/module economics—particularly JKS—while favoring grid automation, storage, inverters and high-voltage transmission suppliers such as Sungrow (300274 CH), CATL (300750 CH) and NARI Technology (600406 CH). AI-linked load growth is unlikely to absorb surplus generation on a one-for-one basis over the next 12 months: data centers require firm, locationally matched power, not merely annual renewable output.

Over 1-3 months, the relevant catalysts are China’s next power-market reform measures, provincial storage mandates, and evidence that utilization rates improve rather than merely capacity additions rise. The more consequential 6-18 month effect is margin migration from panel manufacturing toward dispatch software, battery storage, grid equipment and power-intensive industrial load. This also creates a relative advantage for Chinese AI/data-center operators with access to contracted power, while operators reliant on coal-backed Southeast Asian capacity face higher energy-cost volatility and potentially lower ESG-driven customer demand.

Contrarian risk: the market may overestimate storage as an immediate cure for renewable overbuild. If compensation mechanisms remain weak, batteries can be deployed faster than economically profitable arbitrage opportunities, recreating the oversupply problem downstream. The thesis is falsified if Chinese provincial power prices stabilize and solar utilization rises without a material increase in storage/grid capex, which would indicate the existing system has more absorption capacity than assumed.

Key Decisions for Investors

• Maintain a 6-12 month pair: long CATL (300750 CH) or Sungrow (300274 CH) / short JKS. The trade expresses value migration from commoditized generation equipment to flexibility assets; target 15-25% relative return. Exit if JKS achieves two consecutive quarters of module ASP stabilization and gross-margin expansion while storage pricing deteriorates.
• Add NARI Technology (600406 CH) on pullbacks ahead of Chinese grid-capex and power-market policy announcements. Transmission, dispatch and digital-grid spending is less exposed to solar module pricing; use a 10% downside stop from entry, as delayed State Grid procurement would impair the 2026 earnings catalyst.
• Do not initiate a broad long in Chinese solar manufacturers solely on AI-power-demand narratives. Set an alert for sustained improvement in provincial renewable utilization, battery project IRRs, and data-center power procurement volumes; without those datapoints, additional capacity is more likely to compress industry returns than lift earnings.
• For global portfolios, underweight coal-dependent Southeast Asian data-center and industrial-power exposure versus firms with contracted renewable or gas-backed supply. The risk is a widening delivered-power-cost differential rather than a near-term shortage, with customer procurement and carbon-accounting pressure emerging over 6-18 months.