What happened?
The US Department of Energy's Brookhaven National Laboratory (BNL), working with the Princeton Plasma Physics Laboratory, the DIII-D National Fusion Facility and Colorado-based Great Sky, has developed a neural chip built on superconducting circuits. BNL said the chip mimics human brain networks and can process roughly 1 million fusion plasma samples per second.
Why it matters
Fusion reactors confine plasma at 100-150 million degrees Celsius using magnetic fields. Plasma conditions change rapidly, making real-time monitoring essential. Conventional silicon-based systems struggle to process such volumes fast enough, while superconducting circuits can handle signals at very high speeds.
What the tests showed
The team ran two tests using data from the DIII-D tokamak in California; the chip detected unusual plasma activity and predicted disruptions. In a third experiment, it analyzed simulated plasma data from TORAX, a tokamak core transport simulator developed by Google DeepMind's fusion team, and predicted the onset of sawtooth instabilities.
What's next
Great Sky says the device is its largest superconducting neural chip to date. The team is developing a new chip with a neural network 10 times larger than the current design. The hardware is set to be presented at BNL in early 2027, followed by testing at the DIII-D tokamak later that year. The company says similar chips could find use in quantum computing, cybersecurity and high-speed X-ray data analysis.
