NVIDIA's open Earth-2 weather models are being used by AXA, TotalEnergies and others for sharper local forecasts. Here is what that could mean for business and risk.
Several European companies and universities are now building on NVIDIA's Earth-2, a set of open AI weather models that anyone can adapt with their own data. The newest piece, StormScope EU, predicts cloud conditions across Europe and refreshes every 10 minutes. The goal is better local forecasts for energy, insurance and transport.
A global forecast can show a storm forming. It cannot always say what will happen at one solar farm or along one shipping route. Earth-2 aims to close that gap. Open means developers can take the models and train them on their own information.
StormScope EU was trained on European weather and climate data. It uses satellite observations from EUMETSAT, the European weather satellite agency, across 16 Meteosat channels. At GTC Berlin, NVIDIA is showing a live demo with a company called Earthmover. Visitors can explore the predictions on a map without downloading huge datasets.
Each of these teams started from the same open models and added data that only they have. That is the main pitch of the project.
University of Manchester researchers adapted an Earth-2 model called CorrDiff, which adds local detail to coarser weather data, for air quality in the U.K. They trained it on the Isambard-AI supercomputer. Then they ran it on NVIDIA DGX Spark systems, which sit at a desk. That moves this kind of environmental modeling from a national supercomputer to a single workstation.
The material does not say how accurate these forecasts are compared with current services. It also gives no results from the TotalEnergies tests, and no details on cost or on which uses are already live. The AXA work is a study of risk, not a promise of lower prices. For now, these are early projects with a clear direction.
Developers can try Earth-2 models and Earth2Studio forecasting examples, which NVIDIA has published. If local forecasts keep getting sharper and cheaper to run, firms that depend on weather could plan power, routes and coverage with more detail than they have today.
Open AI weather models are letting energy, insurance and research teams build sharper local forecasts, but accuracy and real-world results are still unproven.
Visa tested 22 AI models on realistic shopping tasks. They passed just over half, and often bought the wrong thing anyway.
Read more: Visa tested 22 AI shoppers, and most still make costly mistakesDARPA wants AI to hunt for design flaws in stock, bond, energy and cloud markets, so enemies cannot use them to cause economic damage that reaches households.
Read more: Pentagon wants AI to find weak spots in stock and energy marketsTuring Award winner Richard Sutton says more computing power alone will not make AI truly smart. Big AI companies disagree, which matters for anyone betting on AI.
Read more: AI pioneer says piling on computing power won't make machines smart
