Nvidia Supercomputers Win Gordon Bell Prize for Climate AI

Two research collaborations using Nvidia-powered supercomputers have been recognised with Gordon Bell Prizes, highlighting how computational capacity is transforming the scientific method itself.

The awards, announced at SC25, the international conference for high-performance computing, recognised work on tsunami forecasting and climate modelling that could reshape how organisations approach disaster response and environmental planning.

Real-time tsunami prediction systems

The University of Texas at Austin, Lawrence Livermore National Laboratory and the University of California San Diego won the Gordon Bell Prize for developing a digital twin that issues real-time probabilistic tsunami forecasts based on a full-physics model. Applied to the Cascadia subduction zone in the Pacific Northwest, the system combines real-time sensor data with full-physics modelling and uncertainty quantification.

Tasks that would normally require 50 years on 512 graphics processing units (GPUs) were completed in just 0.2 seconds on the Alps and Perlmutter supercomputers.

Omar Ghattas, Professor of Mechanical Engineering at the University of Texas at Austin, says: "For the first time, real-time sensor data can be rapidly combined with full-physics modelling and uncertainty quantification to give people a chance to act before disaster strikes."

Omar notes the framework "provides a basis for predictive, physics-based emergency-response systems across various hazards".

Kilometre-scale global climate modelling

A second team from the Max Planck Institute for Meteorology, German Climate Computing Centre, Swiss National Supercomputing Centre, Jülich Supercomputing Centre, ETH Zurich, the University of Hamburg and Nvidia won the Gordon Bell Prize for Climate Modelling for their work on the ICON Earth system model.

The ICON model simulates the entire Earth's systems at kilometre-scale resolution, capturing the flow of energy, water and carbon through the atmosphere, oceans and land. Running on Jupiter, Europe's first exascale supercomputer capable of performing a quintillion calculations per second and hosted at Jülich Supercomputing Centre, the model achieved what researchers describe as a world record in global climate simulation.

The system can simulate approximately 146 days every 24 hours, enabling climate projections decades forward with efficiency that traditional models cannot match.

Daniel Klocke, Computational Infrastructure and Model Development Group Leader at Max Planck Institute for Meteorology, says integrating all components of the Earth system in ICON "at a resolution of 1 kilometre allows researchers to see full global Earth system information on local scales and learn more about the implications of future warming for both people and ecosystems".

Hardware infrastructure enabling discoveries

The research teams used three supercomputers for their work. Alps, hosted at the Swiss National Supercomputing Centre, is powered by more than 10,000 Nvidia GH200 Grace Hopper Superchips, which combine a central processing unit with a graphics processing unit on a single module. Perlmutter, hosted at the National Energy Research Scientific Computing Center in California, runs on Nvidia accelerated computing hardware.

Thomas Schulthess, Director of the Swiss National Supercomputing Centre, says: "At CSCS, we don't just support open science – we accelerate it."

Thomas adds that the work by this year's five Gordon Bell finalists "stand as irrefutable proof: without the Alps supercomputer, these scientific discoveries simply would not exist".

Additional computational breakthroughs recognised

Five teams were shortlisted for the prizes, with projects spanning climate modelling, materials science, fluid simulation, geophysics and electronic design.

Oak Ridge National Laboratory and Nvidia developed ORBIT-2, an AI foundation model for weather and climate downscaling that creates high-resolution data from lower-resolution sources, enabling teams to capture localised phenomena, including urban heat islands and extreme precipitation events.

ETH Zurich developed QuaTrEx, a package of algorithms for nanoscale electronic device modelling. Running on Alps with Nvidia GH200 Superchips, the system can simulate devices with more than 45,000 atoms using 64-bit floating-point precision.

The Georgia Institute of Technology, working with Nvidia, also developed MFC, an open-source solver for fluid flow simulation used in spacecraft design. Running on Alps, the system enables simulation four times faster and with over five times greater energy efficiency than the previous benchmark.

Spencer Bryngelson, Assistant Professor in Computational Science and Engineering at the Georgia Institute of Technology, says the team's "new information geometric regularisation method, combined with the Nvidia GH200 Superchip's unified virtual memory and mixed-precision capabilities, has substantially improved the efficiency of simulating complex computational fluid flows, enabling us to simulate rocket engine plumes at unprecedented scales".