Google's WeatherNext 3 brings hourly, hyperlocal forecasts — and a new edge for renewable energy
Google DeepMind and Google Research have launched WeatherNext 3, an AI weather model that updates forecasts every hour at a resolution five times sharper than its predecessor. Where conventional systems take six hours to process data and output predictions on a 25 km grid, WeatherNext 3 works on a 5 km grid using live satellite imagery — meaning rapidly shifting storms and weather fronts get tracked in something close to real time.
WeatherNext 3 processes live geostationary satellite data to generate hourly forecasts at 5 km resolution.
The upgrade
The jump from WeatherNext 2 is substantial. The old model ran on a 25 km / 6-hour cadence; the new one pulls directly from geostationary satellites, skipping the delay baked into traditional numerical weather prediction (NWP) systems. Precipitation forecasts — rain and snow — are up to 50% more accurate, trained on NASA's IMERG rainfall dataset and satellite radar feeds. The model also factors in ground-level weather station data to account for local terrain, which matters in mountainous or coastal regions where microclimates diverge sharply from regional averages.
The model was trained on NASA IMERG rainfall data and satellite radar feeds, improving precipitation accuracy by up to 50%.
The renewable energy play
The feature that sets WeatherNext 3 apart from general-purpose forecast tools is its energy sector capability. It predicts wind speeds at 100 meters above ground — turbine height — and models cloud cover and solar radiation to help operators estimate output from solar farms. That puts it in direct competition with specialist providers like Solcast and Meteomatics, which US renewable operators currently rely on. Independent benchmarks from Brightband's Operational WeatherBench rank WeatherNext 3 at the top globally, per Gizmodo. For context, NOAA itself deployed a new generation of AI-driven weather models in 2025–26, so operational AI forecasting is quickly becoming the industry standard rather than the exception.
Where it shows up
WeatherNext 3 is already integrated into Google Search, the Gemini app, Google Maps, and Google Earth Engine. Developers and businesses can access hourly weather data via the Maps Platform Weather API and BigQuery. Google still directs users to their local meteorological agency — NOAA and the National Weather Service in the US — for official forecasts, which sidesteps any regulatory friction around replacing government weather infrastructure.
For everyday use, the improvements will filter through quietly: more accurate rain alerts in Google Maps, better storm timing in Search. For energy companies and grid operators, the 100 m wind and solar radiation data via open API access represents a meaningful shift in what's available off the shelf.