NVIDIA Transforms Climate Tech with 'Earth-2': The World's First Fully Open Accelerated AI Climate Stack

For decades, weather forecasting has been the specialty of government supercomputers that run complex physics-based calculations. NVIDIA broke that barrier with the release of the World-2 is a family of open models and tools of AI weather and climate forecasting that is accessible to almost anyone, from tech startups to national weather agencies.

In a move to democratize climate science, NVIDIA unveiled 3 new ground-up models powered by novel architectures: Atlas, StormScopeagain HealDA. These tools promise to accelerate prediction speeds by orders of magnitude while delivering accuracies that rival or exceed conventional methods.

The Democratization of Weather Intelligence

Historically, running a reliable climate model required an infrastructure that only a few countries could afford. NVIDIA's Earth-2 is revolutionizing calculus by offering an 'open stack', a collection of pre-trained models, inference libraries, and customization recipes available on platforms like GitHub and Face Hugging.

Mike Pritchard, Director of Weather Simulation at NVIDIA, stressed that NVIDIA is not a weather service provider. Instead, they create “building blocks” that allow countries and companies to build their own independent forecasting systems.

“Sovereignty matters. Climate is a national security issue… That's why we built Earth-2, the world's first fully open, production-ready AI climate stack.” – Mike Pritchard, NVIDIA

Meet the new Heavyweights: Atlas, StormScope, and HealDA

This announcement introduces 3 specific models that address different stages of the forecasting pipeline, from processing raw data to predicting storms weeks in advance.

1. Earth-2 Medium Range (Powered by Atlas)

Targeting a 15-day forecast window, this model uses a new structure called Atlas. It predicts more than 70 weather conditions, including wind, humidity, and pressure, with high accuracy.

• Performance: In industry standard benchmarks, the Atlas has been shown to perform very well GenCastthe best open model of the moment, there are many variations.
• Shift: It represents a return to “simple, unruly Transformer architecture,” moving away from niche, hand-crafted AI designs.

2. Earth-2 Nowcasting (Powered by StormScope)

This is a game changer for rapid disaster response. Powered by StormScopethis generative AI model focuses on a 0- to 6-hour window, providing kilometer-scale resolution of local storms.

• Why it matters: It is the first AI model to outperform traditional physics-based methods for forecasting short-term rainfall.
• Speed: It produces severe weather forecasts in minutes, giving emergency responders valuable time to act.
• Sovereignty: Because it is trained directly on geostationary satellite images rather than region-specific physics results, it can be deployed by any country with a good satellite.

3. Earth-2 Global Data Assimilation (Powered by HealDA)

Often the unsung hero of forecasting, “data assimilation” is the process of combining dirty satellite and balloon data into a coherent summary of the atmosphere to start a forecast.

• Success: Traditional simulation consumes about 50% of supercomputing cycles. NVIDIA HealDA architecture accomplishes this task by minutes on GPUs there are hours on mainframes.
• Result: When combined with the Medium Range model, it produces the most skillful predictions ever seen in a fully AI-based pipeline.

Real World Impact: From Solar Energy to Hurricane Risk

The Earth-2 stack is already being used by the world's major players, proving that AI weather forecasting is ready for commercialization and prime time.

• Renewable Energy: Absolute Power again GCL (a major manufacturer of solar panels) use Earth-2 to predict solar and wind fluctuations. For solar farms, accurate prediction of cloud cover can have a significant impact on energy market transactions.
• Israel weather service: Using the CorrDiff model (part of the Earth-2 family), they found a 90% reduction in calculation time while producing high-resolution forecasts up to eight times a day.
• Insurance and Risk: AXA again S&P Global Energy they use the speed of Earth-2 to run thousands of “unreal” scenarios. By simulating thousands of years of hypothetical storm data, they can better understand rare, high-impact weather events that haven't happened yet but could happen.
• Daily Activities: Light beltan AI weather tool provider, already integrates Earth-2 Medium Range to issue daily global forecasts.

The Bottom Line

NVIDIA Earth-2 isn't just a technology upgrade; it is a structural change in the way people interact with the climate. By lowering the barrier to entry, from multi-million dollar supercomputers to GPU-accelerated AI, NVIDIA is enabling a future where highly accurate weather forecasting is ubiquitous.

As extreme weather events become more frequent, so do the tools StormScope again Atlas it will likely become critical infrastructure for governments and industries around the world.

Earth-2 Medium Range and Nowcasting are available on GitHub, Hugging Face, and NVIDIA Earth2Studio. Earth-2 Global Data Assimilation is expected to be released later this year.

To learn more about getting started with these models, developers can visit the NVIDIA Earth-2 technical blog. Earth-2 Medium Range [Read the research paper], Earth-2 Nowcasting [Read the research paper]again Earth-2 Global Data Assimilation [Read the research paper].

Jean-marc is a successful AI business executive .He leads and accelerates the development of AI-powered solutions and started a computer vision company in 2006. He is a well-known speaker at AI conferences and has an MBA from Stanford.