An international collaboration led by Google DeepMind in the UK developed an AI weather model called “WeatherNext Cyclone (WN-C)” that simultaneously predicts the track, size and intensity of tropical cyclones, with results published in Nature on 6 August. Trained on several decades of global atmospheric reanalysis data from ECMWF plus a dataset of about 5,000 tropical cyclones, WN-C probabilistically generates more than 50 possible scenarios up to 15 days ahead. Evaluated on 2023–2025 tropical cyclones, it secured on average more than one extra day of forecast lead time over major operational models for track, maximum wind speed and the radius of gale-force winds. Mean track error at five days was about 230 km, 140 km smaller than ECMWF’s 370 km, reaching accuracy roughly 30 hours earlier. For three-day maximum wind speed, WN-C was on average 3.75 knots more accurate than NOAA’s HAFS. Using about 0.25-degree data, it outperformed high-resolution regional models on intensity, improved rapid intensification balance, and as a 1,000-member ensemble captured extreme wind risks. Added to NHC consensus forecasts, track errors fell an additional 28% and intensity errors 6%. Limitations include no direct rainfall, storm surge or localized gust prediction, and dependence on high-quality initial analyses. Trial predictions have been provided to NHC forecasters since the 2025 Atlantic hurricane