Gordon Bell Prize research continues to advance global science by enabling faster simulations, more accurate models and greater computational efficiency. Within the first 20 words, Gordon Bell Prize research appears naturally to set the theme for this work. Scientists now rely on powerful AI supercomputers to study climate dynamics, electronic design, geophysics and fluid behavior. As computing power grows, teams can run full-physics simulations that once required months. Now, they finish them in hours or seconds. This progress allows researchers to ask deeper questions and validate complex ideas more confidently. Additionally, open science practices help broaden access, since many finalists publish their results for global use. These shared findings strengthen collaboration and support new breakthroughs. With supercomputers like Alps, JUPITER and Perlmutter powering these efforts, researchers can model Earth systems at kilometer resolution, simulate nanoscale devices and create foundation models that improve scientific understanding.
Breakthrough Digital Twins in Gordon Bell Prize Research
Two winning teams at SC25 demonstrated major progress in digital twin technology. The first created the world’s initial real-time tsunami digital twin based on a full-physics model. Developed by the University of Texas at Austin, Lawrence Livermore National Laboratory and the University of California San Diego, the system generates probabilistic forecasts within fractions of a second. Normally, these simulations would take decades. However, on Alps and Perlmutter they run at remarkable speed. This capability gives coastal regions a greater chance to respond before disaster strikes. Furthermore, the model blends real-time sensor data with uncertainty quantification, showing how physics-based forecasting can support emergency planning across multiple hazards.
Read Also
NVIDIA Materials Discovery Breakthrough
Accelerated AI Storage Solutions
Climate Modeling Breakthroughs Powered by Gordon Bell Prize Research
Another winning team earned the Climate Modelling Prize for its work on the ICON Earth system model. The project involved researchers from the Max Planck Institute for Meteorology, DKRZ, CSCS, JSC, ETH Zurich, the University of Hamburg and NVIDIA. ICON models Earth at kilometer-scale resolution, allowing energy, water and carbon cycles to be simulated with exceptional detail. Running on the JUPITER supercomputer, ICON set a world record in global climate simulation. It compresses roughly 146 simulated days into 24 hours. This speed supports long-term climate projections with improved precision. Additionally, the project uses CUDA Graphs and CUDA-X libraries to enhance performance. These tools help scientists analyze extreme rainfall events, local climate impacts and atmospheric behavior more effectively.
Advances in Electronics, Fluid Dynamics and Weather Modeling
Several other finalist projects delivered important results. At ETH Zurich, the QuaTrEx project advanced nanoscale electronic device modeling. Using the Alps supercomputer, researchers simulated devices with more than 45,000 atoms at FP64 precision. This helps accelerate the design of next-generation transistors. In fluid dynamics, the Georgia Institute of Technology developed MFC, an open-source solver that runs simulations four times faster and with higher energy efficiency. It maintains accuracy while scaling across thousands of GPUs. Full-scale runs on JUPITER are expected to break previous records and support more detailed spacecraft design. ORBIT-2, created by Oak Ridge National Laboratory, NVIDIA and partners, advances weather and climate modeling through spatial hyper-resolution downscaling. Running on Alps, it predicts local climate shifts such as extreme rainfall, monsoon changes and urban heat with improved clarity. These achievements show how Gordon Bell Prize research blends physics, AI and computation to drive scientific progress.
