Google says its AI weather model is getting better

Google has announced the launch of WeatherNext 3, its newest artificial intelligence weather model aimed at significantly improving the accuracy and speed of global forecasts.

While traditional meteorological techniques depend heavily on supercomputers executing complex physics equations to simulate atmospheric conditions, this updated model leverages real-time satellite data to make rapid predictions. By learning from these fresher, richer observational datasets, the system can produce a global picture that is five times sharper than its predecessor. It is capable of updating hourly and mapping environmental variables like temperature and moisture at a highly localized level, which provides substantial improvements for predicting rain and snowfall.

These advancements are particularly valuable in regions outside the United States and Europe that lack extensive ground-based monitoring equipment like rain gauges. In addition to powering updates across Google's consumer ecosystem—including Maps, Search, and Gemini—the technology is tailored to predict wind speeds at heights matching turbine blades, thereby supporting renewable energy grid planning. This development arrives as tech companies face pressure to address the massive energy footprint of their data centers. Despite these technological leaps, Google's researchers note that the artificial intelligence model works in tandem with traditional physical simulation techniques rather than replacing them entirely.