New AI Beats Supercomputers in Weather Forecasting with Drastically Lower Energy Use

Aardvark Weather is a cutting-edge AI-driven forecasting system that delivers predictions tens of times faster than traditional supercomputer methods while using thousands of times less energy. Developed in just 18 months, it harnesses raw data from satellites, weather stations, ships, and balloons to generate rapid forecasts on a desktop computer. Its streamlined, end-to-end machine learning approach reduces reliance on complex physics models, promising cost reductions and greater accessibility, especially for regions with limited computational resources.

In early 2025, Aardvark Weather made headlines by matching or exceeding the performance of established forecasting systems. The model can be specifically tailored for sectors like agriculture and renewable energy, and its flexibility hints at broader applications across Earth system forecasting. Supported by renowned institutions and research collaborations, this breakthrough heralds a significant step forward in making advanced weather prediction more available and energy-efficient globally.