NVIDIA and General Atomics have made a major leap forward in the pursuit of commercial fusion energy by developing a high-fidelity, AI-powered digital twin of a fusion reactor. The project, unveiled at the NVIDIA GTC Washington, D.C. conference, leverages advanced AI models, supercomputing resources, and the NVIDIA Omniverse platform to create an interactive virtual environment that mirrors the real-world DIII-D National Fusion Facility operated by General Atomics. The digital twin integrates data from physics-based simulations, engineering models, real-time sensor inputs, and AI surrogate models trained on decades of fusion research. This unified system runs on NVIDIA GPUs and is powered by the company’s CUDA-X libraries, enabling rapid, accurate predictions of plasma behavior—something that once took weeks on supercomputers now happens in seconds. Key AI models developed for the project include EFIT for plasma equilibrium, CAKE for plasma boundary prediction, and ION ORB for modeling heat density from escaping ions. These models allow researchers to monitor and stabilize the plasma in real time, reducing the risk of damage and accelerating experimental cycles. The project was supported by major scientific computing centers: the Argonne Leadership Computing Facility (ALCF), the National Energy Research Scientific Computing Center (NERSC), and the San Diego Supercomputer Center at UC San Diego. Supercomputers such as Polaris at ALCF and Perlmutter at NERSC were used to train the AI models at scale. Plasma, the fourth state of matter and the fuel of fusion, operates at temperatures exceeding hundreds of millions of degrees Celsius—conditions that make controlling it an immense scientific challenge. By using AI to simulate and predict plasma dynamics with unprecedented speed and accuracy, researchers can now test countless scenarios virtually before conducting physical experiments. The digital twin serves as a collaborative platform for a global team of 700 scientists from over 100 institutions. It enables real-time interaction, rapid prototyping, and the ability to run “what-if” experiments without risking the physical reactor. This not only increases safety but also dramatically shortens the timeline for innovation. The project marks a shift in fusion research, moving from a purely physics-driven process to one enhanced by AI, data science, and high-performance computing. By turning weeks-long simulations into near-instantaneous insights, the digital twin acts as a powerful accelerator for achieving practical fusion energy. This collaboration underscores how AI is becoming essential to solving some of humanity’s most complex scientific challenges. With continued advances, the dream of harnessing the power of the sun on Earth is becoming increasingly within reach.