Nvidia is pushing artificial intelligence deeper into climate science. On Monday, the chipmaker introduced a new suite of Nvidia AI weather forecasting models designed to dramatically speed up weather predictions while reducing costs—without sacrificing accuracy.

The announcement was made at the American Meteorological Society’s annual meeting in Houston and builds on Nvidia’s broader strategy of pairing open-source AI software with its powerful computing hardware.

Replacing Traditional Weather Simulations With AI

Weather forecasting has long relied on physics-based simulations that are both computationally expensive and time-consuming. Nvidia says its AI weather forecasting approach can rival—and in some cases exceed—the accuracy of these traditional models once training is complete.

Crucially, the models are far cheaper and faster to run. That speed advantage opens the door to new use cases that were previously impractical due to cost and compute limits.

Why Speed Matters: Insurance and Extreme Events

One of the most immediate applications of Nvidia AI weather forecasting lies in the insurance industry. Insurers need to understand rare but catastrophic events—such as floods, hurricanes, or severe storms—to price risk accurately.

Historically, this has been difficult because weather forecasts are run in “ensembles,” meaning thousands of slightly different simulations must be calculated to capture extreme outliers. Each high-resolution simulation is slow, making large ensembles prohibitively expensive.

Mike Pritchard, Nvidia’s director of climate simulation research and a professor at the University of California, Irvine, says AI changes that equation entirely.

“Once trained, AI is 1,000 times faster,” Pritchard said. “So you’re free to run massive ensembles. Insurance companies are now running 10,000-member ensembles.”

Inside Nvidia’s Earth-2 AI Models

The newly released models fall under Nvidia’s Earth-2 initiative and include three distinct systems:

• A model designed for 15-day global weather forecasts
• A short-range model focused on severe U.S. storms up to six hours ahead
• A data-fusion model that integrates information from multiple weather sensors to improve downstream forecasting tools

Together, they aim to cover both long-range planning and short-term, high-impact weather events.

Open Source, Hardware-Powered Strategy

Nvidia emphasized that the models are open source, reinforcing its push to make AI tools widely accessible—while still benefiting from the company’s dominance in AI hardware.

This mirrors Nvidia’s broader playbook across industries, from generative AI and robotics to autonomous driving: provide powerful, open software that runs best on Nvidia chips.

A Bigger Shift in Climate and AI

The launch of Nvidia AI weather forecasting models highlights a growing shift in climate science. Instead of relying solely on physics-based simulations, researchers are increasingly blending them with AI-driven systems that can explore far more scenarios in less time.

If widely adopted, these models could reshape how governments, insurers, and industries prepare for extreme weather—making forecasts not just faster, but more actionable.

Don’t miss out on our latest news—follow us for the latest AI news, breakthroughs, and insights that matter.