International Edition
Latest News
World

China Uses 100,000 Domestic AI Chips for Global Weather Forecast System

Chinese researchers have successfully generated a comprehensive 10-day global weather forecast in approximately one hour using an artificial intelligence supercluster powered by 100,000 domestic AI chips, according to the South China Morning Post. Developed jointly by the China…

China Uses 100,000 Domestic AI Chips for Global Weather Forecast System

Chinese researchers have successfully generated a comprehensive 10-day global weather forecast in approximately one hour using an artificial intelligence supercluster powered by 100,000 domestic AI chips, according to the South China Morning Post. Developed jointly by the China Meteorological Administration and the computing firm Sugon, the system—known as Sugon 8000—operates a homegrown meteorological model to process planetary atmospheric data at a 5-kilometer grid resolution.

Sugon 8000 Architecture and Global Weather Modeling Performance

The Sugon 8000 system links 100,000 high-performance computing devices to execute high-resolution global projections. Traditional meteorological systems typically run on standard supercomputers that prioritize precision physics simulations, yet the Sugon AI cluster sacrifices minor margins of precision to process massive volumes of atmospheric data at high speeds. During recent benchmark testing, the system produced a 10-day global forecast in roughly 60 minutes. While this calculation speed sits close to the roughly 45-minute runtime of the Global Forecast System operated by the United States National Weather Service, Sugon representatives noted that a one-hour delivery window satisfies the operational requirements for daily meteorological forecasting.

Grid Resolution Standards and Regional Forecasting Goals

Current operational weather forecasting models worldwide generally feature grid intervals between 5 and 10 kilometers. The Sugon 8000 platform achieves a 5-kilometer grid spacing globally, while internal testing has already validated capabilities down to a 3-kilometer grid resolution. Higher grid resolutions directly improve the detection of localized hazardous weather events such as flash floods and heavy snowfall, although the computational cost scales exponentially. Halving the grid interval can increase required computing performance by up to eightfold. The Chinese government maintains a stated target to deploy a nationwide 1-kilometer grid resolution weather forecasting network by 2030, with plans to refine resolution down to 100 meters in critical metropolitan and geographical zones.

About the author: Ibrahim Khalil - World Editor

PhD in International Relations, former UN press officer. Ibrahim has reported from 40+ countries, translating complex geopolitical shifts into clear, human‑focused narratives. “Ibrahim Khalil provides authoritative world news, from diplomacy to conflict zones, with on‑the‑ground insight.”