Upgraded system targets storms and local extremes
Google’s DeepMind has rolled out a new version of its AI-powered weather prediction model that aims to deliver more precise forecasts, especially for short-notice storms and localized extreme events. The upgraded system, known as WeatherNext 2, uses vast amounts of satellite, radar and station data to predict how temperature, wind and rainfall will evolve over the next few hours and days. Engadget reports that the model is designed to outperform traditional numerical weather prediction tools on key measures like heavy rainfall and storm tracks.

National meteorological agencies in Europe and parts of Asia have already been testing earlier versions of DeepMind’s approach. The new release expands coverage and adds tools tailored for grid operators, airlines and emergency planners. By running on specialized AI hardware in Google’s data centers, WeatherNext 2 can generate updated forecasts more frequently, which is crucial when storms intensify faster than expected. Researchers say the system has shown particular promise in “nowcasting”—very short-term predictions vital for flash-flood warnings and urban drainage planning.

Climate resilience and data-center questions
The launch comes as climate change makes extreme weather more frequent and harder to predict using older models alone. Heatwaves, cloudbursts and cyclones are stressing infrastructure worldwide, from European rivers to South Asian coasts. AI-enhanced forecasts could help authorities in countries like Bangladesh evacuate at-risk communities earlier, adjust dam releases and protect crops ahead of sudden downpours. Insurers and reinsurers are also eyeing such tools to refine their risk models as payouts from weather-related disasters climb.

At the same time, the push for AI forecasting raises questions about energy use and control over critical data. Training and running large models consumes significant electricity, and much of today’s AI computing still depends on power grids that burn fossil fuels. DeepMind says WeatherNext 2 is more efficient than previous versions, but campaigners argue that climate tech should be transparent about its own footprint. There is also concern that too much reliance on proprietary systems from a handful of tech giants could weaken public weather services, especially in poorer countries.

Meteorologists stress that AI will complement, not replace, existing physics-based models and human forecasters. The best results so far have come from hybrid systems that blend traditional simulations with machine-learning tools. Still, the rapid progress of AI weather models suggests that emergency planners, farmers and city authorities will soon have access to more granular, neighborhood-level forecasts. The challenge will be ensuring those insights reach vulnerable communities in time—and that the benefits are shared beyond the wealthiest markets.