This post was originally published on this site.
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.
Techcrunch event
San Francisco
|
October 13-15, 2026
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.”
