Weather forecasting is getting a serious upgrade, and Nvidia Corp. (NVDA) wants to make sure everyone can play. At the American Meteorological Society's Annual Meeting, the chipmaker unveiled its Earth-2 family of open models, libraries, and frameworks, billing it as the world's first fully open, accelerated weather AI software stack. The goal? Help more organizations build faster, cheaper prediction systems without needing a supercomputer the size of a warehouse.
The pitch is straightforward. Traditional weather forecasting requires massive computational resources and time. Nvidia's Earth-2 tools promise to speed every stage of the process, from crunching raw observation data to producing 15-day global forecasts or hyper-local storm predictions. And because the platform is open, it theoretically lowers the barrier to entry for smaller agencies and companies that want sophisticated forecasting without the traditional infrastructure costs.
What's Actually In The Toolbox
Earth-2 Medium Range, built on something called the Atlas architecture, can predict weather up to 15 days ahead across more than 70 variables. That's your classic extended forecast capability, but turbocharged by AI.
Then there's Earth-2 Nowcasting, powered by StormScope, which uses generative AI trained on satellite and radar data to produce zero-to-six-hour storm forecasts in minutes. This is the "what's happening right now and very soon" tool, critical for emergency management and operations planning.
Perhaps most interesting is Earth-2 Global Data Assimilation, powered by HealDA. This tool generates atmospheric "initial conditions" in seconds on GPUs, a process that traditionally takes hours on supercomputers. Nvidia says pairing this with Medium Range creates the most accurate predictions available from an open, all-AI pipeline. Initial conditions matter because they're the starting point for any forecast model, and getting them right and fast is half the battle.
Who's Already Using It
Nvidia reports that developers and agencies are already running and testing Earth-2 tools. The list includes Brightband, the Israel Meteorological Service, Taiwan's Central Weather Administration, The Weather Company, and the U.S. National Weather Service. On the commercial side, energy and finance players like TotalEnergies, Eni, GCL, Southwest Power Pool (working with Hitachi), AXA, and S&P Global Energy are kicking the tires.
Earth-2 Medium Range and Nowcasting are available now through Nvidia Earth2Studio and on Hugging Face and GitHub. Earth-2 Global Data Assimilation will launch later this year.
The Bigger Picture
Nvidia's move comes as AI-powered weather forecasting heats up. Last July, Palantir Technologies Inc. (PLTR) expanded its decision-making software into AI-powered weather forecasting through a partnership with Tomorrow.io. Palantir teamed up with the weather analytics firm to embed real-time atmospheric data and predictive weather tools directly into its platforms, targeting defense, aviation, government, and infrastructure customers who need better operational readiness.
The question is whether open platforms like Earth-2 can deliver enough accuracy and customization to compete with proprietary systems, and whether organizations have the technical chops to implement them effectively. But if the adoption list is any indication, there's real appetite for alternatives to traditional supercomputer-dependent forecasting.
Price Action: Nvidia (NVDA) shares were down 0.43% at $186.86 at the time of publication on Monday.
