Key developments: Leading agencies are now running AI alongside traditional supercomputers, with some AI runs matching or beating the Met Office’s massive computers (WeatherQuest). ECMWF’s AI ensemble can produce 51 scenarios more than ten times faster, using roughly 1,000 times less energy and improving accuracy up to 20%. Google DeepMind’s WeatherNext 2 delivers 15‑day forecasts eight times faster, giving cyclone tracks a day‑early warning. The U.S. National Hurricane Center has begun basing its forecasts for storms such as Isaias on DeepMind models, citing greater consistency and earlier intensity warnings. Academic work shows AI can anticipate rare, extreme events—like the April 2024 Dubai flooding—five days in advance, while a Luxembourg flood study confirms AI excels at light rain and temperature forecasts but still trails physics‑based models for the heaviest rains that trigger flash floods.
Why it matters: Faster, energy‑efficient AI runs lower operational costs and enable more ensemble members, delivering probabilistic forecasts that improve risk‑based decision‑making for energy, insurance and emergency services. Accuracy gains—up to 20% on some metrics—mean better day‑to‑day weather guidance and earlier storm warnings, directly protecting lives and property. Demonstrated ability to predict unprecedented extreme events addresses a critical gap in climate‑resilience planning.
What to watch next:
1. Wider operational rollout of ECMWF’s AI ensemble and Google’s WeatherNext 3, expanding the number of ensemble members and resolution.
2. Continued integration of AI guidance into national hurricane centres and other agencies, especially for rapid‑intensification forecasts.
3. Research on AI performance for heavy‑rain, flash‑flood scenarios, aiming to close the remaining gap with physics‑based models.
4. Growth of commercial AI‑driven services as market forecasts anticipate a $4.70 billion sector by 2034, driven by faster, cheaper AI forecasts.
These trends suggest AI weather forecasting is moving from experimental prototypes to core operational tools, reshaping how societies prepare for both everyday weather and climate‑driven extremes. [[1, 2]]