Google DeepMind Releases WeatherNext 2 — AI Weather Model Partnered with US National Hurricane Center for Cyclone Forecasting

Following its Google I/O debut, Google Research and DeepMind released WeatherNext 2 for broad deployment — a significantly improved AI weather forecasting model with 8x faster generation than its predecessor and 1-hour temporal resolution. In a landmark for AI-powered disaster preparedness, WeatherNext was formally partnered with the US National Hurricane Center (NHC) to provide AI-powered tropical cyclone track predictions alongside traditional ensemble forecasts. The model is available via Google Earth Engine and BigQuery for government agencies and researchers. WeatherNext 2 also powers Google's Urban Flash Flood forecasting system — providing 24-hour advance warnings for urban flash floods in 150+ countries via Flood Hub — using a recurrent LSTM neural network trained on 2.6 million historical flood incidents sourced via the Groundsource AI (powered by Gemini). The NHC partnership represents a milestone: official adoption by the US federal meteorological agency responsible for protecting the Caribbean, Gulf of Mexico, and Atlantic hurricane-prone populations. Research published alongside the release shows 12-hour AI lead time reduces flash flood damage by up to 60%.