Brain-Inspired Superconducting Chip Keeps Pace With Fusion Plasma at One Million Samples per Second

Published in Scientific American on October 22, 2026, the research describes a collaboration between Brookhaven National Laboratory, AI hardware startup Great Sky, Princeton Plasma Physics Laboratory, and the DIII-D National Fusion Facility in California. The team developed a superconducting neuromorphic chip — circuitry that mimics the brain's parallel processing architecture using analog superconducting loops rather than conventional silicon transistors — and tested it against recorded plasma data from the DIII-D tokamak replayed at a rate of one million samples per second. The chip successfully classified plasma anomalies and predicted imminent disruptions — sudden collapses of the superhot plasma that can release enormous energy and physically damage a reactor's inner wall — in real time without missing a beat. Researchers emphasise that the work is still at a demonstration stage and the chip has not yet been connected to a live reactor; however, a successor chip with a neural network ten times larger is planned for testing at Brookhaven in early 2027, ahead of potential installation at DIII-D. If scaled successfully, the technology could become a critical safety layer for future commercial fusion plants.