Gordon Bell Prize finalists are redefining the frontiers of open science through groundbreaking applications of AI and high-performance computing powered by NVIDIA technologies. Among the standout projects is ORBIT-2, an exascale AI foundation model for weather and climate modeling developed through a collaboration between Oak Ridge National Laboratory, NVIDIA, and other partners. Running on the Alps supercomputer, ORBIT-2 leverages exascale computing and innovative algorithms to deliver unprecedented scalability and precision in climate downscaling—transforming coarse-resolution climate data into hyper-local, high-resolution forecasts. This advancement enables scientists to model intricate regional phenomena such as urban heat islands, extreme rainfall events, and subtle shifts in monsoon systems with far greater accuracy than traditional models. According to Prasanna Balaprakash, director of AI programs and section head for data and AI systems at Oak Ridge National Laboratory, NVIDIA’s advanced supercomputing platforms were instrumental in achieving exceptional performance, reliability, and scientific impact. Another finalist project, a digital twin for tsunami early warning, marks a transformative leap in disaster preparedness. Developed by researchers from the University of Texas at Austin, Lawrence Livermore National Laboratory, and the University of California San Diego, this system delivers real-time probabilistic tsunami forecasts using a full-physics model. When applied to the Cascadia subduction zone in the Pacific Northwest, it performed computations equivalent to 50 years of simulation in just 0.2 seconds—achieving a 10 billion-fold speedup on the Alps and Perlmutter supercomputers. “This is the first time real-time sensor data can be fused with full-physics modeling and uncertainty quantification to provide actionable warnings before a disaster strikes,” said Omar Ghattas, professor of mechanical engineering at UT Austin. He emphasized that the framework lays the foundation for predictive, physics-based emergency response systems applicable to a wide range of natural hazards. Key to the performance of both the tsunami digital twin and the ICON and MFC projects were NVIDIA CUDA-X libraries, which optimized simulation efficiency and scalability. The ICON project further benefited from NVIDIA CUDA Graphs, enabling complex workflows to be defined as computational graphs rather than sequential operations—boosting performance and reducing latency. These achievements exemplify how AI and exascale computing are accelerating scientific discovery and empowering open science. Researchers across disciplines are now able to tackle problems once considered computationally intractable. For those interested in the latest advancements in supercomputing, NVIDIA will be showcasing these innovations at SC25, taking place through Thursday, November 20.