As of September 2026: NVIDIA Earth-2 Use Cases Across Europe Span Weather, Energy, Insurance and Remote Sensing

What happened

As of September 2026, NVIDIA describes Earth-2 as an end-to-end weather and climate AI software stack made up of open models, libraries and frameworks for research institutions, startups and governments to run, fine-tune and deploy forecasting systems on their own infrastructure. Earth-2 models can combine StormScope for short-term forecasts, CorrDiff for added local detail and PhysicsNeMo for training and customization with proprietary data and domain expertise.

As of September 2026, StormScope EU is Earth-2’s European regional model. It uses European weather and climate data plus EUMETSAT observations across 16 Meteosat channels, updated every 10 minutes, to predict cloud evolution. NVIDIA and Earthmover demonstrated a live pipeline in which Earth2Studio produces forecasts, Earthmover’s platform provides the results as map tiles and updates them every 10 minutes when new satellite observations arrive.

Who is affected

As of September 2026, AXA is using FourCastNet 3 to study regional cyclone risk beyond the historical record by simulating more than 1,000 seasons, supporting risk assessment and prevention planning. TotalEnergies is evaluating StormScope-based real-time solar-radiation forecasts at two photovoltaic sites and developing subhourly weather forecasts to help anticipate changes in solar-power generation.

As of September 2026, University of Manchester researchers adapted CorrDiff for U.K. air-quality modeling, trained it on the Isambard-AI supercomputer and ran it locally on NVIDIA DGX Spark systems. Vaisala Xweather combines Earth-2 with its own observations, lightning data and forecasting expertise for government, energy, maritime, transportation and insurance customers, adapting StormScope for software-as-a-service products and AI-assistant integration. Space42 developed SARDiff4Height with CorrDiff and NVIDIA PhysicsNeMo to estimate surface height from a single radar image and produce uncertainty maps.

The case for and against

As of September 2026, NVIDIA says Earth-2 models can combine StormScope for short-term forecasts, CorrDiff for added local detail and PhysicsNeMo for training and customization with proprietary data and domain expertise. NVIDIA’s listed cases span cyclone risk, solar forecasting, air quality and remote sensing.

As of September 2026, TotalEnergies’ work includes evaluating real-time forecasts at two photovoltaic sites and developing subhourly weather forecasts, while AXA’s case is a simulation of more than 1,000 seasons to study regional cyclone risk. The available material does not specify formal commercial scale or actual decision-making outcomes for these cases.

Risks

As of September 2026, the evidence does not provide revenue, contract values or incremental NVIDIA revenue for the cases. It also does not provide independent comparisons of accuracy, latency or cost against traditional physical models or other AI models. The deployment stages, formal commercialization timelines and actual decision-making outcomes for TotalEnergies, AXA and the other cases are not specified.

What to watch

As of September 2026, watch whether TotalEnergies discloses results from its evaluation at two photovoltaic sites and how AXA’s more than 1,000 simulated seasons are used for risk assessment and prevention planning. Also watch for independent disclosures on performance, cost, latency, commercialization timelines or revenue; the evidence does not provide an attributable relationship between NVIDIA’s share price and the Earth-2 cases.