StormScope EU brings 10-minute AI nowcasts to Europe’s insurers and energy firms

Nvidia added StormScope EU to Earth-2, its family of open AI weather and climate models and tools. The short-term forecasting model is trained on European data and uses EUMETSAT satellite observations to predict evolving cloud conditions across 16 Meteosat channels at 10-minute intervals. At GTC Berlin, Nvidia showed a live StormScope EU demo built with Earthmover that refreshes as new satellite observations arrive every 10 minutes. Earth-2 also includes CorrDiff, which adds local detail to coarse data, and the PhysicsNeMo framework for customization.

Global forecasts show the storm; decisions about solar output, shipping and insured property depend on what happens over one valley in the next hour. That local, short-range layer is where these tools are aimed, and where the first adopters, insurers, energy producers and forecasters, have money at stake. It is part of a broader GTC Berlin slate built around open models that run near the data (Cosmos world models).

AXA is using FourCastNet 3 to study cyclone risk across more than 1,000 simulated seasons, beyond the historical record, valuable for understanding tail risk. Vaisala Xweather is adapting StormScope with its lightning data, showing incumbents blending open models with proprietary observations rather than being displaced. TotalEnergies is evaluating solar nowcasts at two photovoltaic sites. Space42 produces uncertainty maps alongside its radar-based height estimates. University of Manchester researchers adapted CorrDiff for U.K. air-quality modeling, trained it on the Isambard-AI supercomputer, then ran it locally on desktop-class DGX Spark machines, moving environmental modeling from national infrastructure to a lab bench (DGX Spark clusters).

Publishing these as open models that organizations can retrain with their own data is a different pattern from a single black-box forecast sold by subscription, and widens who can do this work.

Most of what is described is still exploration, and the proof will be seasons of verified forecasts compared with operational European services. AI forecasts can look confident where they are weakest, so uncertainty estimates like Space42’s should become standard output rather than a research extra.

The insurance case also carries a consumer dimension. Simulated seasons beyond the historical record are how catastrophe and pricing models get built; if AI-generated scenarios begin to drive premiums, the assumptions inside those simulations matter to policyholders, and insurers will face questions about disclosure. Nvidia has framed related climate work as both a clean-energy and a grid story (Nvidia AI for clean energy). More in our Nvidia coverage and research coverage.