NVIDIA has unveiled the Earth-2 family of open models—the world’s first fully open, accelerated set of AI-driven tools designed for weather forecasting. These models are poised to transform how weather predictions are made, enabling faster, more accurate forecasts that support critical decisions in agriculture, energy, public health, and climate resilience. The Earth-2 suite includes models like Earth-2 Medium Range, Earth-2 CorrDiff, and Earth-2 Nowcasting, all built to run efficiently on NVIDIA’s AI infrastructure. By making these models open source, NVIDIA is empowering researchers, weather agencies, climate-tech innovators, and enterprises to deploy, fine-tune, and extend them on their own local AI systems, accelerating innovation across the global weather and climate community. Brightband, a weather AI provider and member of NVIDIA’s Inception program under the Sustainable Futures initiative, is already using Earth-2 Medium Range to deliver daily global forecasts. “The revolution in AI-powered weather tools is accelerating rapidly, and models like Earth-2 Medium Range are at the forefront,” said Julian Green, cofounder and CEO of Brightband. “As one of the first organizations to run this model operationally, we’re seeing how open access drives faster iteration, collaboration, and improvement across the weather enterprise.” The Israel Meteorological Service is using Earth-2 CorrDiff in daily operations and plans to integrate Earth-2 Nowcasting to generate high-resolution forecasts up to eight times per day. “NVIDIA Earth-2 models reduce compute time by 90% at 2.5-kilometer resolution compared to traditional numerical weather prediction on CPU clusters,” said Amir Givati, director of the Israel Meteorological Service. “After a recent storm, our AI model trained with CorrDiff outperformed all other operational models in predicting six-hour accumulated rainfall.” The Weather Company is evaluating Earth-2 Nowcasting for localized severe weather applications, while the U.S. National Weather Service is assessing the models to enhance its operational forecasting workflows. In the energy sector, TotalEnergies is exploring Earth-2 Nowcasting to strengthen short-term risk awareness and decision-making. “Earth-2 represents a major leap in operationalizing advanced weather intelligence at scale,” said Emmanuel Le Borgne, climate and weather forecast product manager at TotalEnergies. “For energy systems where timing and local conditions are critical, these models deliver transformative value.” Eni is actively testing Earth-2 models, including FourCastNet and CorrDiff, for semi-operational downscaling to generate probabilistic, high-resolution forecasts of weather and gas demand up to weeks in advance. GCL, a leading Chinese solar materials producer and global energy operator, has deployed Earth-2 models in production for its photovoltaic forecasting system. Compared to traditional numerical methods, Earth-2 delivers more accurate predictions at significantly lower cost, greatly improving the precision of solar power generation forecasts. Southwest Power Pool, working with Hitachi, is using Earth-2 Nowcasting and FourCastNet3 to enhance intraday and day-ahead wind forecasting. This advancement supports the region’s grid reliability goals and enables more informed, real-time operational decisions. On the financial side, S&P Global Energy is leveraging Earth-2 CorrDiff to convert climate data into actionable local insights for risk assessment. Meanwhile, AXA, a global insurance group, is using FourCastNet to simulate thousands of hypothetical hurricane scenarios as part of its research into model evaluation, methodology development, and benchmarking against existing techniques. By combining open access with high-performance AI, NVIDIA’s Earth-2 initiative is setting a new standard for how weather intelligence is developed, shared, and applied—driving progress across industries and helping societies better prepare for a changing climate.