Nvidia dropped three open-source AI weather models on Monday. The announcement came at the American Meteorological Society’s annual meeting in Houston.
These aren’t just incremental improvements. The company is taking direct aim at conventional weather forecasting methods.
Traditional weather simulations eat up time and money. They require massive computational resources to generate predictions.
Nvidia’s AI-driven approach changes that equation. The models deliver comparable or better accuracy while running faster and cheaper.
The speed difference is striking. Once trained, these AI models run 1,000 times faster than conventional methods.
Mike Pritchard leads climate simulation research at Nvidia. He also teaches earth system sciences at UC Irvine.
Pritchard pointed to insurance companies as prime users. These firms need to understand extreme weather events.
Think massive floods. Think hurricanes that could devastate entire regions.
The traditional approach to predicting extreme events involves “ensembles.” These are groups of individual weather predictions starting from the same point.
Finding outlier events requires running many ensemble members. More members mean better coverage of possible scenarios.
But here’s the problem. Calculating each member in precise detail takes time.
Want to know if a specific property will flood? That requires detailed calculations for each ensemble member.
The AI solution removes that bottleneck. Insurance companies can now run 10,000-member ensembles without breaking a sweat.
Pritchard calls the old tension “gone.” The speed advantage opens new possibilities for risk assessment.
The “Earth-2” suite includes three distinct models. Each serves a specific forecasting purpose.
The first model handles 15-day weather forecasts. This covers the medium-range prediction window.
The second specializes in severe storms over the United States. It focuses on forecasts up to six hours out.
The third model tackles a different problem. It integrates data from various weather sensors.
Weather data comes from many sources. Satellites, ground stations, weather balloons, and ocean buoys all contribute.
Getting these disparate streams to work together is challenging. The integration model creates better starting points for other forecasting technology.
Nvidia released all three models as open-source software. Any organization can access and use them.
This fits the company’s broader strategy. They’re providing open-source software for various AI applications powered by their chips.
The applications range from chatbots to self-driving vehicles. Weather forecasting joins that growing list.
The models were announced Monday, January 26, 2026. They’re available now through Nvidia’s Earth-2 platform.
The post Nvidia (NVDA) Stock: Company Launches Free AI Models for Weather Prediction appeared first on CoinCentral.
Also read: Blizzard : des annonces importantes arrivent sur Overwatch, Diablo et WoW
