Is Your AI Fast Enough to Run a Fusion Reactor?

Abstract

In magnetic confinement fusion, compact machine learning algorithms are often integrated into the control loop on CPU-based custom hardware stacks. Deterministic implementations of these algorithms target instabilities which can evolve over microseconds to milliseconds. Staying within this framework may therefore limit the pace of control-system development to the inference time available. With the expansion of hardware acceleration and large models, we aim to bring ML models for fusion control to modern inference engines and compare, based on the physics reactor, the suitability of each engine per model. This study implements a consistent real-time machine learning algorithm for GPUs and optimized CPU packages, including OpenVINO, XLA, and TVM, at DIII-D National Fusion Facility. For the smallest model, one CPU core responds within 3.6 μs in 99% of control cycles and an NVIDIA A100 within 28 μs. We also evaluate models running simultaneously, as they would run in a typical experiment, and find best use cases per framework.

Fast Cliffords When Your Quantum Memory Is Full
Marten Folkertsma, Ian Mertz, Sergii Strelchuk +1
arXiv30 Sep 2026PreprintarXiv licenceFull textQuantum physics & quantum informationarXiv
Managing complexity for integrated fusion reactor design using a differentiable graph-based MDAO framework: PROCESS as a case study
Timo J. Bogaarts, Willem Rutten, Felix Warmer
arXiv15 Oct 2026PreprintCC BY 4.0Full textComputational physicsarXiv
Diffusion model for fast alpha-particle transport in Magneto-Inertial Fusion plasmas
Alexander Velikovich, Fernando García-Rubio, Scott Davidson +3
arXiv28 Oct 2026PreprintarXiv licenceFull textPlasma physicsarXiv
Direct Production of Ultra-High Radioisotopic Pure ⁶⁴Cu from Natural Zinc Metal via D-D Fusion Narrowband Fast Neutron Source Irradiation
Jeremy M. C. Brown, Mantas Grigalavicius, Tibor Gilinger +7
arXiv22 Oct 2026PreprintCC BY-NC-ND 4.0Full textBiological & medical physicsarXiv
Generative modeling for fusion: a large-scale benchmark of stellarator fast-ion simulations
Misha Padidar, Teresa Huang, Mark Goldstein +2
arXiv27 Oct 2026PreprintCC BY 4.0Full textPlasma physicsarXiv

Cite

BibTeX RIS

BibTeX

@misc{chen2026your,
  title = {{Is Your AI Fast Enough to Run a Fusion Reactor?}},
  author = {Nathaniel Chen and Andrew Rothstein and Ricardo Shousha and Hiro Farre-Kaga and Peter Steiner and Azarakhsh Jalalvand and Egemen Kolemen},
  howpublished = {arXiv},
  year = {2026},
  month = oct,
  doi = {10.48550/arxiv.2610.00845},
  eprint = {2610.00845},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2610.00845},
  note = {CC BY 4.0},
}

RIS

TY  - UNPB
TI  - Is Your AI Fast Enough to Run a Fusion Reactor?
AU  - Nathaniel Chen
AU  - Andrew Rothstein
AU  - Ricardo Shousha
AU  - Hiro Farre-Kaga
AU  - Peter Steiner
AU  - Azarakhsh Jalalvand
AU  - Egemen Kolemen
PY  - 2026
DA  - 2026/10/01
PB  - arXiv
DO  - 10.48550/arxiv.2610.00845
UR  - https://arxiv.org/abs/2610.00845
AB  - In magnetic confinement fusion, compact machine learning algorithms are often integrated into the control loop on CPU-based custom hardware stacks. Deterministic implementations of these algorithms target instabilities which can evolve over microseconds to milliseconds. Staying within this framework may therefore limit the pace of control-system development to the inference time available. With the expansion of hardware acceleration and large models, we aim to bring ML models for fusion control to modern inference engines and compare, based on the physics reactor, the suitability of each engine per model. This study implements a consistent real-time machine learning algorithm for GPUs and optimized CPU packages, including OpenVINO, XLA, and TVM, at DIII-D National Fusion Facility. For the smallest model, one CPU core responds within 3.6 μs in 99% of control cycles and an NVIDIA A100 within 28 μs. We also evaluate models running simultaneously, as they would run in a typical experiment, and find best use cases per framework.
ER  -

Abstract

Related

Cite