How NVIDIA’s Open AI Predicts Weather and Climate

Accurate weather forecasting has always been important – from protecting communities from extreme weather to optimising renewable energy generation.

Now, rapid advances in AI are transforming how we understand and predict the planet’s dynamic atmosphere. 

AI-powered weather and climate models can process vast amounts of observational data in real time, deliver localised forecasts in minutes and reduce the energy and cost footprint of traditional supercomputing systems. This democratisation of high-quality climate insight is unlocking opportunities for scientists, enterprises and policymakers to act faster – and smarter – in a warming world.

Against this backdrop, NVIDIA has unveiled Earth‑2, a major leap forward in open-access weather and climate AI. 

“Making production-ready weather AI fully accessible for organisations to run, fine-tune and deploy on their own infrastructure, NVIDIA Earth-2 is the first open stack to bring together typically disparate weather AI capabilities, from generating current atmospheric conditions to predicting weather weeks in advance,” says Mike Pritchard, Director of Climate Simulation at NVIDIA.

“It’s pioneering work to speed weather prediction, enhance forecasting accuracy, foster collaboration and advance scientists’ overall understanding of the planet’s atmospheric  conditions.”

Earth‑2 introduces the world’s first fully open, accelerated weather AI software stack – combining pretrained models, frameworks, customisation recipes and inference libraries.

Its goal: to bring powerful, production-ready forecasting tools to anyone working toward greater planetary resilience.

A unified platform for climate intelligence

Traditional weather forecasting depends on physics-based models running on energy-intensive supercomputers. 

While powerful, these models are slow and costly, limiting their availability to only a few large national institutions. 

NVIDIA Earth‑2 replaces that bottleneck with a modular AI infrastructure capable of accelerating every stage of the forecasting process – from data assimilation to medium- and short-term predictions. By deploying these models on GPUs instead of traditional CPU clusters, users can cut processing times from hours to seconds.

Built as an open ecosystem, Earth‑2 allows researchers, developers, and public agencies to run, fine-tune, and deploy AI weather models directly on their own infrastructure. The platform unifies several architectures within one framework, fostering scientific collaboration and deeper understanding of atmospheric behavior. 

As a result, more organisations – from weather services to energy traders – can generate bespoke, high-resolution forecasts adapted to their operational needs.

NVIDIA’s family of open AI models

The Earth‑2 portfolio now spans multiple model architectures tackling different layers of the weather system. Earth‑2 Medium Range, built on NVIDIA’s new Atlas model, delivers 15-day forecasts across more than 70 weather variables, outperforming leading open-source competitors in accuracy. 

Earth‑2 Nowcasting, powered by StormScope, uses generative AI to produce kilometer-scale, zero- to six-hour predictions of local storms and hazardous weather – generating actionable results within minutes.

Complementing these is Earth‑2 Global Data Assimilation, underpinned by a model called HealDA, which produces atmospheric “snapshots” – initial conditions essential for accurate prediction. When coupled with the Medium Range model, it creates a fully AI-powered pipeline capable of delivering some of the most skilful forecasts yet recorded.

These join earlier Earth‑2 components, including CorrDiff, which downscales coarse global data to fine local resolutions up to 500 times faster than traditional methods, and FourCastNet3, a high-accuracy model producing 60x faster forecasts across variables like wind, temperature, and humidity. 

Together, these open models create an integrated foundation for a new era of weather intelligence.

Real-world adoption and sustainability impact

Earth‑2’s open access has already drawn diverse collaborators. Brightband, an AI weather tool provider and member of NVIDIA’s Sustainable Futures initiative, uses Earth‑2 Medium Range operationally to issue daily forecasts.

The Israel Meteorological Service credits the CorrDiff model with reducing compute requirements by 90% while improving accuracy for precipitation verification.

“NVIDIA Earth-2 models give us a 90% reduction in compute time at 2.5-kilometer  resolution compared with running a classic numerical weather prediction model without AI on a CPU cluster,” adds Amir Givati, Director of the Israel Meteorological Service. “After a recent rainstorm, our AI model trained with CorrDiff was the best of all our operational  models for a six-hour verification of accumulated precipitation.” 

In the energy sector, TotalEnergies, Eni and GCL are leveraging Nowcasting and FourCastNet models to enhance short‑term risk awareness and optimise solar and gas forecasts. Meanwhile, AXA and S&P Global Energy are using Earth‑2’s generative models to simulate thousands of extreme-weather scenarios, advancing climate risk assessment and insurance modelling.

“NVIDIA Earth-2 represents a major step forward in how advanced weather intelligence  can be operationalized at scale,” said Emmanuel Le Borgne, Climate and Weather Forecast Product Manager at TotalEnergies. 

“Models like Earth-2 Nowcasting are ground-breaking for our business because they improve short-term risk awareness and decision-making in  energy systems where minutes and local impacts matter.” 

All sustainability, net zero and sustainable supply chain leaders should attend:

• Sustainability LIVE: The Net Zero Summit - QEII Centre, London, March 4-5
• Sustainability LIVE: The US Summit - Navy Pier, Chicago, April 21-22

Co-located with Procurement & Supply Chain LIVE, these events brings together CSOs, ESG leaders and senior decision-makers at a moment when sustainability, supply chains and commercial performance are increasingly interconnected.

Tickets can be booked online today for The Net Zero Summit and The US Summit. Group discounts available.

Supporting a safer climate future

By making its models openly available on NVIDIA Earth2Studio, GitHub and Hugging Face, NVIDIA is positioning Earth‑2 as a catalyst for global collaboration in climate technology.

Alongside initiatives like NVIDIA PhysicsNeMo – which lets developers train hybrid physics‑AI systems – Earth‑2 represents a new frontier where open science and AI innovation meet to enhance planetary resilience.

As extreme weather grows more frequent and costly, tools like Earth‑2 demonstrate the dual potential of AI: accelerating scientific progress while supporting a sustainable global future. In empowering governments, industries and innovators to see further and act faster, NVIDIA’s Earth‑2 may help rewrite not just the forecast – but the future itself.