US launches STELLAR-AI to slash nuclear fusion simulation time from months to real-time

STELLAR-AI Transforms Fusion Energy Research with AI-Powered Computing

The Princeton Plasma Physics Laboratory (PPPL) has launched STELLAR-AI, a computing platform designed to eliminate the computational delays that have long plagued fusion energy research1

1

. The Simulation, Technology, and Experiment Leveraging Learning-Accelerated Research enabled by AI initiative pairs artificial intelligence with high-performance computing to reduce simulation time from months to real-time analysis2

2

. This shift addresses a critical challenge: fusion research involves complex simulations of plasma behavior that currently require several months to complete using existing infrastructure, creating simulation bottlenecks that slow progress toward commercial fusion power.

Real-Time Data Analysis Connects Computing to Experiments

STELLAR-AI connects computing resources directly to experimental devices, allowing researchers to analyze data as experiments occur rather than waiting months for simulation results2

2

. A primary experimental connection is the National Spherical Torus Experiment-Upgrade (NSTX-U) at PPPL, scheduled to begin operations this year1

1

. "Fusion is a complex system of systems. We need AI and high performance computing to really optimize the design for economic construction and operation," said Jonathan Menard, deputy director for research at PPPL1

1

. This real-time capability represents a fundamental shift in how fusion scientists can test hypotheses and refine designs.

Advanced Hardware Architecture Powers AI-Driven Simulations

The hardware architecture combines central processing units (CPUs), graphics processing units (GPUs), and quantum processing units (QPUs) in a configuration tailored to fusion challenges1

1

. CPUs handle standard computing tasks, while GPUs provide the parallel processing capabilities necessary for training AI models. QPUs solve specific complex calculations that traditional computers cannot process efficiently1

1

. This setup meets the technical needs of private companies developing commercial fusion power solutions and supports the creation of digital twins of fusion devices like NSTX-U—computer models that mirror physical machines so closely that scientists can test ideas virtually before running actual experiments2

2

.

Genesis Mission and National Fusion Strategy

STELLAR-AI operates as part of the Genesis Mission, a national effort launched by executive order in November 2025 to use AI to accelerate scientific discovery across DOE laboratories2

2

. The platform contributes fusion-specific computer codes and scientific models to this broader national system while supporting the DOE's Fusion Science and Technology Roadmap1

1

. The roadmap outlines development of an AI-Fusion Digital Convergence platform aimed at commercialization of fusion power plants and providing energy for future computing and AI infrastructure. One project called StellFoundry uses AI to speed the design of stellarators, a type of fusion device with a twisted, pretzel-like shape that some scientists believe could offer advantages over other designs1

1

.

Industry and Academic Collaboration Accelerates Development

The project brings together national laboratories, universities, technology companies, and private fusion firms to build the computational foundation the fusion community needs2

2

. Partner laboratories include the UK Atomic Energy Authority (UKAEA), while academic participation includes Princeton University, the Massachusetts Institute of Technology, and the University of Wisconsin-Madison1

1

. NVIDIA is working to improve the performance of fusion-related computer codes, and Microsoft provides cloud integration through its Azure service1

1

. Private industry partners include Commonwealth Fusion Systems, General Atomics, Type One Energy, and Realta Fusion, positioning the platform to provide tools and AI models for the U.S. fusion industry as companies race to bring solutions to market1

1

. This coalition signals that the path to commercially viable fusion power depends on computational speed paired with precision, fundamentally changing the timeline for fusion energy development.