China's AI data centers confront green‑power shortfall

Beijing has set an ambitious target that 80% of the electricity used by AI data centres must come from renewable sources by 2030. In 2023, renewable power accounted for only about 11% of data‑centre consumption, and analysts estimate that demand could rise by 3,000–5,000 billion kWh between 2026 and 2030 – roughly the annual electricity use of a large industrial province.

Experts say the gap is technical, not political. AI training workloads create rapid, high‑amplitude power peaks, while inference requires a continuous baseline load. The existing grid struggles to absorb such intermittent renewable input without voltage fluctuations that can damage expensive GPU hardware. As a result, many operators prefer stable coal‑based supply, sometimes supplementing the shortfall with green‑energy certificates that do not guarantee physical delivery.

Chinese firms are developing “AI power systems” that integrate weather‑forecasting models, storage, and millisecond‑level grid control to match renewable output with AI demand. Envision Energy showcased a pilot in Inner Mongolia where a 100% renewable power supply is linked directly to a Tencent AI data centre, and it plans larger green AI hubs elsewhere. The success of such systems will determine whether China can meet its climate pledge while sustaining the rapid expansion of AI computing.