Google AI is a hurricane genius and is here to stay for storm predicting

The National Hurricane Center (NHC) began using Google DeepMind’s AI model operationally in June 2025, marking a major shift in hurricane forecasting. During the 2025 season, which featured three Category 5 storms—the most since 2005—the AI system delivered predictions meteorologists called stunning, achieving a decade’s worth of advancement in tropical cyclone forecasting. It operates 100 times faster than traditional physics-based models, requiring no supercomputer, and adapts by learning on the fly. AccuWeather senior hurricane forecaster Alex DaSilva highlighted the AI’s superiority, noting it outperformed the NHC in early storm predictions and will be used more extensively in future seasons. The NHC’s 2025 performance report confirmed DeepMind’s accuracy, with no high-profile forecast failures in the Atlantic. Michael Lowry, a hurricane specialist at WPLG TV in Miami, described it as consistently the best guidance across the Eastern Pacific and North Atlantic basins. Despite its success, meteorologists criticized the model as a ‘black box’ due to its lack of transparency. Forecasters cannot explain why it performs well or access details like wind shear, storm structure, or rainfall data—key factors for predicting freshwater flooding, a deadly hazard in recent years. John Cangialosi, an NHC senior hurricane specialist, acknowledged the AI’s value but warned it does not replace traditional models entirely. The AI’s rapid processing and self-learning capabilities allow for earlier hurricane watches and warnings, giving communities more time to prepare. However, its limitations—such as missing rainfall insights—highlight the need for hybrid approaches combining AI with human expertise. Developers, primarily software engineers, continue refining the model to address these gaps while leveraging its unmatched predictive power for storm tracking.