AI inference speed limits fusion reactor control systems
Is Your AI Fast Enough to Run a Fusion Reactor?
Performance
Summary
Controlling nuclear fusion reactors requires very fast and predictable AI decisions. The authors show that large machine learning models run much slower on CPUs than on GPUs, which can affect control performance. They tested different AI models used in fusion plants and created a benchmark to compare their speeds. Their findings highlight the need to choose both the AI model and the hardware together to meet critical timing requirements.
What this means in practice
- •For fusion control engineers: Select AI hardware and models to ensure timely feedback control in fusion reactors.
- •For real-time system developers: Optimize AI inference deployments by matching model size with hardware speed to meet strict timing constraints.
Authors
Nathaniel Chen, Andrew Rothstein, Ricardo Shousha, Hiro Farre-Kaga, Peter Steiner, Azarakhsh Jalalvand, Egemen Kolemen
Abstract
Machine learning models are increasingly used in feedback control loops for nuclear fusion, where inference speed and predictable timing are critical. We summarize lessons from models deployed for control on the DIII-D tokamak and develop a benchmark to compare inference backends across ten neural networks and model components from fusion control and diagnostic pipelines. For models greater than five million parameters, the CPU backends take tens to thousands of milliseconds, while GPU inference is substantially faster, suggesting an upper limit on CPU-oriented development for control. These results show why the deployment backend must be selected together with the model and its control-cycle budget.