Domestic AI Supercomputer Enables Breakthrough in Global Kilometer-Scale Weather Forecasting

In September 2026, the Earth System Numerical Prediction Center of the China Meteorological Administration, together with Sugon, announced a key breakthrough in global kilometer-scale numerical weather forecasting. Using the Sugon 8000, a fully domestic AI supercomputing cluster with 100,000 GPUs, and the independently developed MCV numerical model, the team completed a global 5-kilometer-resolution, 10-day weather forecast within one hour. The operational efficiency met international mainstream standards and passed professional forecast quality verification.

High-resolution global weather simulation consumes enormous computing resources, and the difficulty grows exponentially with finer resolution and broader coverage. The joint effort combined meteorological research with supercomputing engineering, optimizing algorithms and system architecture in both directions. After achieving stable 5-kilometer resolution, the team also completed a global 3-kilometer simulation experiment, laying the technical foundation for future higher-precision operational forecasting.

The achievement marks a new stage in the integration of domestic meteorological forecasting and supercomputing power. According to the team, the breakthrough also provides a replicable model for high-intensity simulation fields such as life sciences, new material research, and engineering simulation, showing that domestic supercomputing infrastructure is extending from hardware deployment to supporting major scientific research and frontier innovation.