US researchers have built a brain-like computer chip that can analyze fusion plasma at one million samples per second. Fusion reactors heat plasma to temperatures roughly 10 times hotter than the Sun’s core. As a result, monitoring these extreme conditions requires processing power that conventional silicon chips struggle to deliver.
Brookhaven National Laboratory designed the chip in partnership with Colorado-based startup Great Sky. The company manufactured it based on 12 years of research at the National Institute of Standards and Technology. In addition, the team worked with Princeton Plasma Physics Laboratory (PPPL) and the DIII-D National Fusion Facility in California. Together, they tested the chip’s ability to monitor and predict plasma behavior in real time.
“This is our largest chip to date, and it’s great to see it processing high-speed data directly from a tokamak,” said Jeff Shainline, Great Sky’s founder and CEO. “This is just the beginning. This approach to AI hardware has so much to offer with regard to fusion, as well as other applications at the forefront of technology. Analysis of high-speed X-ray data, advanced cyber security, control and error correction of quantum systems, video intelligence with continual learning — these are all incredible opportunities as we scale up our systems.”
The chip uses superconducting circuits arranged to mimic neurons in the human brain. Instead of running on silicon transistors like traditional processors, it operates at cryogenic temperatures. It also communicates through light signals down to single-photon levels. Consequently, it can process massive data streams with far less energy, according to Great Sky.
In lab tests, the team fed real plasma data from the DIII-D tokamak into the chip at one million samples per second. The chip successfully detected unusual plasma behavior and predicted disruptions that could damage a reactor. Furthermore, a third experiment used simulated data from TORAX, a tokamak simulator built by Google DeepMind’s fusion team. The chip then predicted sawtooth instabilities, a common challenge in tokamak operations.
Great Sky plans to unveil a next-generation chip with a neural network 10 times larger in December 2026. The team also aims to test it during live fusion experiments at DIII-D in 2027. Beyond fusion, the company says similar chips could support quantum computing, cybersecurity, and high-speed video analysis.