China Bets on AI Weather Forecasting as Extreme Storms Close In
China is placing its bets on Artificial Intelligence (AI) for weather forecasting as extreme storms pose increasing threats. Three Chinese models are making significant strides, narrowing the gap with Western labs, with one reportedly predicting a typhoon's landfall within half an hour.
August 11, 2026 - 2:30 pm
Credit: Canva
As Typhoon Dolphin approached China's coast, the country's investment in AI-driven weather forecasting became more than just academic research; it became critical infrastructure. With extreme weather events intensifying and the cost of inaccurate forecasts rising, Beijing is allocating substantial resources to develop models that can provide early and cost-effective storm predictions.
Three notable Chinese systems lead this effort:
- Fengwu from Shanghai AI Laboratory
- Pangu developed by Huawei
- Fuxi created at Fudan University
This diverse group—comprising a state lab, a tech giant, and a leading university—is racing to establish AI supremacy in forecasting, a field with real-world implications.
The claims are impressive: Fengwu reportedly outperformed Google DeepMind's GraphCast across 80% of weather variables tested, extending useful forecasts beyond the traditional ten-day mark. If these results hold up to independent scrutiny, it indicates that China is rapidly narrowing the gap with Western leaders in this field.
During Typhoon Dolphin, Fengwu demonstrated its capabilities by predicting the storm's impact location and time just 30 minutes and 30 kilometers ahead, thanks to Sun Zhi of Techwind. This level of precision enables effective evacuation plans, turning what could be guesswork into a logistical strategy.
AI weather forecasters offer significant advantages over traditional methods: they generate predictions far quicker and require less computational power. For China, with its extensive coastal regions and weather-sensitive agriculture, speed and cost efficiency are paramount.
Sun emphasizes the practical value:
"With more extreme weather, people need information to make decisions, both local governments, the national government, also the average person, farmers and fishermen."
Despite these advancements, Chinese forecasters acknowledge challenges. AI models still struggle with predicting storm intensity, which is crucial for distinguishing manageable events from catastrophic ones. Additionally, they have yet to be fully tested against significant climate changes that fall outside historical data.
However, China's efforts are gaining momentum in a highly competitive global field. They compete with well-funded startups and established players like Nvidia-backed FourCastNet, the European Centre for Medium-Range Weather Forecasts' AIFS, and even a Swiss firm claiming its model surpasses Microsoft and Google.