Nvidia launches Earth-2 open AI weather forecasting models, beating Google's GenCast on accuracy

Nvidia's new AI weather models probably saw this storm coming weeks ago

In the run-up to the winter storm currently pummeling much of the U.S., weather forecasts for some regions were all over the map, with snowfall predictions varying wildly. Nvidia couldn't have timed the release of its new Earth-2 weather forecasting models any better. Or, given how accurate the company claims the new models are, maybe it knew something we didn't? The new AI models promise to make weather forecasting faster and more accurate. Nvidia claims that one model in particular, Earth-2 Medium Range, beats Google DeepMind's AI weather model, GenCast, on more than 70 variables. GenCast, which Google released in December 2024, was itself significantly more accurate than existing weather models that were capable of generating forecasts up to 15 days out. Nvidia announced the new tools Monday at the American Meteorological Society meeting in Houston. "Philosophically, scientifically, it's a return to simplicity," Mike Pritchard, director of climate simulation at Nvidia, told reporters on a call before the meeting. "We're moving away from hand-tailored niche AI architectures and leaning into the future of simple, scalable, transformer architectures." Traditionally, most weather forecasts rely on simulations of physics as observed in the real world. AI models are a relatively recent addition. The Earth-2 Medium Range model is based on a new Nvidia architecture called Atlas, about which the company said it would release more details on Monday. Alongside Medium Range, Nvidia's Earth-2 suite includes a Nowcasting model and Global Data Assimilation model. Nowcasting produces short-term predictions from zero to six hours into the future, and it's aimed at helping meteorologists forecast the impacts of storms and other hazardous weather. "Because this model is trained directly on globally available geostationary satellite observations, rather than region-specific physics model outputs, Nowcasting's approach can be adapted anywhere on the planet with good satellite coverage," Pritchard said. That should help governments of states and smaller countries understand how severe weather systems might affect their territories. The Global Data Assimilation model uses data from sources like weather stations and balloons to produce continuous snapshots of weather conditions at thousands of locations around the world. Those snapshots are then used as launching points for weather models to make their predictions. Traditionally, those snapshots have required tremendous amounts of computing power before the forecasting work could begin. "It consumes roughly 50% of the total supercomputing loads of traditional weather [forecasting]," Pritchard said. "This model can do that in minutes on GPUs instead of hours on supercomputers." The three new models join two existing ones: CorrDiff, which uses coarse-grained forecasts to generate speedy, high-resolution predictions, and FourCastNet3, which models individual weather variables like temperature, wind, and humidity. Pritchard said that the new models should give more users access to powerful weather forecasting tools, which have historically been the domain of wealthier countries and large corporations, which have the funds to pay for costly supercomputer time. "This provides the fundamental building blocks used by everyone in the ecosystem -- national meteorological services, financial service firms, energy companies -- anyone who wants to build and refine weather forecasting models," Pritchard said. Some of the tools are already in use. Meteorologists in Israel and Taiwan have been using Earth-2 CorrDiff, for example, while The Weather Company and Total Energies are evaluating Nowcasting, Nvidia said. "For some users, it makes sense to subscribe to an enterprise centralized weather forecasting system. But for others like countries, sovereignty matters," Pritchard said. "Weather is a national security issue, and sovereignty and weather are inseparable."

NVIDIA Launches Earth-2 Family of Open Models -- the World's First Fully Open, Accelerated Set of Models and Tools for AI Weather

NVIDIA Earth-2 Global Data Assimilation shows the complex patterns of Earth observation data from satellites, weather balloons and weather stations, which the AI model transforms into smooth, continuous estimates of the atmospheric state from which forecasts can be launched. Accurate weather forecasting helps save lives and protect environments -- and is a cornerstone of decision-making in agriculture, energy, public health and other industries. Researchers, weather agencies, climate-tech innovators and enterprises are already running, fine-tuning and building on these state-of-the-art models to unlock scientific breakthroughs using their own local AI infrastructure. Weather Forecasting AI weather tool provider Brightband -- a member of the NVIDIA Inception program's Sustainable Futures initiative -- is running Earth-2 Medium Range to issue real-world global forecasts daily. "The revolution of new AI weather tools for forecasting is very exciting and continues to gather speed with new models like NVIDIA Earth-2 Medium Range," said Julian Green, cofounder and CEO of Brightband. "Brightband is among the first to run Earth-2 Medium Range operationally, and the model being open source speeds up innovation, allowing easier comparison and improvements by other members of the weather enterprise." The Israel Meteorological Service is using Earth-2 CorrDiff in operation -- and plans to use Earth-2 Nowcasting -- to generate high-resolution forecasts up to eight times daily, enabling decision-makers to respond more effectively to extreme weather while reducing computational costs. "NVIDIA Earth-2 models give us a 90% reduction in compute time at 2.5-kilometer resolution compared with running a classic numerical weather prediction model without AI on a CPU cluster," said Amir Givati, director of the Israel Meteorological Service. "After a recent rainstorm, our AI model trained with CorrDiff was the best of all our operational models for a six-hour verification of accumulated precipitation." The Weather Company is evaluating Earth-2 Nowcasting for localized severe-weather applications, and NWS is evaluating the new models to enhance its operational workflows. Energy Forecasting and Grid Operations TotalEnergies is evaluating Earth-2 Nowcasting to improve short-term risk awareness and decision-making. "NVIDIA Earth-2 represents a major step forward in how advanced weather intelligence can be operationalized at scale," said Emmanuel Le Borgne, climate and weather forecast product manager at TotalEnergies. "Models like Earth-2 Nowcasting are groundbreaking for our business because they improve short-term risk awareness and decision-making in energy systems where minutes and local impacts matter." Eni is intensively testing Earth-2 models, including FourCastNet and CorrDiff, for semi-operational downscaling of predictions to produce probabilistic, high-resolution forecasts of weather and gas demand weeks ahead. GCL, one of China's largest solar material producers and a global integrated energy operator, is running NVIDIA Earth-2 models in operation for its photovoltaic prediction system. Compared with traditional numerical weather prediction, Earth-2 provides more accurate prediction data at a lower cost, significantly improving the accuracy of GCL's photovoltaic power generation prediction. Southwest Power Pool, in collaboration with Hitachi, is using Earth-2 Nowcasting and FourCastNet3 to improve intraday and day-ahead wind forecasting. This effort supports Southwest Power Pool's commitment to enhancing grid reliability and enabling more informed operational decisions across the SPP footprint. Financial Impact Assessment S&P Global Energy is harnessing NVIDIA Earth-2 CorrDiff to turn climate data into local insights for risk assessment. Global insurance group AXA is using FourCastNet to generate thousands of hypothetical hurricane scenarios as part of its R&D program in model evaluation, methodological development and benchmarking on existing techniques.

Nvidia launches Earth-2 open AI weather forecast models and tools

Nvidia launches Earth-2 open AI weather forecast models and tools Nvidia Corp. today unveiled Earth-2, its artificial intelligence models and tools for scientists, startups, developers, enterprises and governments worldwide to make weather prediction more accessible than ever. Access to weather prediction matters because it shapes the map of human decisions, from the mundane to the profound. If a logistics leader knows a rainstorm will hit within the next day, disrupting dockworkers, that forecast changes staffing and routing before the first drop falls. If a government official oversees critical infrastructure and sees a polar vortex on the horizon, six feet of snow in four days, then planning isn't optional: it's triage, staging and continuity. The Earth-2 open technologies include pretrained models, frameworks, customization recipes and inference libraries. They allow users to accelerate all forecasting stages from observation data to generating 15-day global or local storm forecasts. Historically, production-ready weather has been the mainstay of supercomputer-based models. Generative AI has broken free of these computation- and cost-intensive modelling simulation needs, making it accessible to startups and nations alike. Included in the toolset is Earth-2 Medium Range, a 15-day high-accuracy predictive model powered by a new model architecture called Atlas for medium-range forecast across more than 70 weather variables, including temperature, pressure, wind and humidity. Earth-2 Nowcasting, powered by a new model called StormScope, which uses generative AI to make country-scale forecasts into kilometer-resolution, zero- to six-hour predictions of local storms and hazardous weather in just minutes. Earth-2 Global Data Assimilation, powered by HealDA, produces initial conditions for weather prediction. These are snapshots of current atmospheric conditions at thousands of locations around the globe. Assimilation can generate initial conditions in seconds using graphics processing units instead of hours on supercomputers. When coupled with Earth-2 models, Nvidia said, this results in accurate forecasting predictions. Earth-2 CorrDiff can downscale continental predictions into high-resolution, regional weather for fine-grain local forecasting. FourCastNet3 can deliver high-accuracy forecasting for various weather variables such as wind, temperature and humidity, surpassing leading conventional models, producing answers at 60 times faster speeds. Weather forecasting partners Numerous partners joined Nvidia to build using Earth-2, including Brightband, an AI weather tool provider, running Medium Range to issue global forecasts daily. "Brightband is among the first to run Earth-2 Medium Range operationally, and the model being open-source speeds up innovation, allowing easier comparison and improvements by other members of the weather enterprise," said Julian Green, co-founder and chief executive. Power companies including Eni S.p.A., GCL Technology Holdings Ltd. and S&P Global Energy, are using Earth-2 tools for predictive weather analysis to operationalize. "Nvidia Earth-2 represents a major step forward in how advanced weather intelligence can be operationalized at scale," said Emmanuel Le Borgne, climate and weather forecast product manager at TotalEnergies SE. Nvidia said the full-featured open-model family and tools for training capabilities for diagnostics will be licensed for commercial and noncommercial use. They will also be available on GitHub and Hugging Face upon release.

Nvidia unveils Earth-2 open models for AI weather forecasting By Investing.com

Investing.com -- Nvidia launched a new family of open models and tools for AI weather forecasting at the American Meteorological Society's Annual Meeting on Monday. The Earth-2 family offers the world's first fully open, accelerated weather AI software stack, making weather and climate prediction more accessible to scientists, startups, developers, enterprises and government agencies globally. The new technology accelerates all forecasting stages, from processing initial observation data to generating 15-day global forecasts or local storm predictions. This AI-powered approach saves significant computational time and costs compared to traditional physics-based models run on supercomputers. The newly announced models include Earth-2 Medium Range (powered by Atlas architecture), Earth-2 Nowcasting (powered by StormScope), and Earth-2 Global Data Assimilation (powered by HealDA). These join existing models in the Earth-2 stack such as Earth-2 CorrDiff and Earth-2 FourCastNet3. Several organizations are already implementing these technologies. AI weather tool provider Brightband is running Earth-2 Medium Range to issue daily global forecasts. The Israel Meteorological Service uses Earth-2 CorrDiff operationally, reporting a 90% reduction in compute time compared to traditional methods. Energy companies including TotalEnergies, Eni, GCL, and Southwest Power Pool are evaluating or implementing Earth-2 models to improve forecasting for their operations. Financial firms like S&P Global Energy and AXA are using the technology for risk assessment and scenario generation. Earth-2 Medium Range and Nowcasting are now available via NVIDIA Earth2Studio, Hugging Face, and GitHub, with Earth-2 Global Data Assimilation expected later this year. This article was generated with the support of AI and reviewed by an editor. For more information see our T&C.