Scientists in New York state have developed a unique superconducting neural chip that can reportedly process about a million fusion plasma samples per second.
The new Super Intelligence (SI) computer chip was developed by researchers at the U.S. Department of Energy’s (DOE) Brookhaven National Laboratory (BNL). For the project, they collaborated with the Princeton Plasma Physics Laboratory (PPPL), the DIII-D National Fusion Facility and the Colorado-based technology company Great Sky.
BNL noted that the chip relies on superconducting circuits to mimic human neural networks. It could eventually improve the way fusion reactors monitor and control hot plasma. Laboratory tests showed that the chip could detect unusual plasma behavior and predict instabilities.
“It is incredible to see the team processing data from our tokamak with this unique chip,” said Richard Buttery, PhD, DIII-D director. The team believes the technology could support faster and more reliable control systems for fusion reactors.
Handling plasma data
Fusion reactors use magnetic fields to maintain hot plasma at temperatures between 100 and 150 degrees Celsius (180 and 270 million degrees Fahrenheit). This extreme temperature is about ten times hotter than the Sun’s core.
Still, plasma conditions can change quickly, making real-time monitoring essential to keep them stable. This requires computers that can process huge amounts of data almost instantly. But traditional silicon-based computer systems struggle to process such large amounts of data quickly enough.
Photo credit: Great Sky Inc.
However, the new superconducting neural chip can process signals at extremely high speeds. It uses superconducting circuits arranged to mimic the connections in the human brain.
To test it, scientists collected data from the DIII-D Tokamak, a General Atomics fusion research facility in California. Two of these tests evaluated the chip’s ability to detect unusual plasma activity and predict disruptions.
The data was collected at a rate of one million samples per second, and the chip successfully handled this pace. “This is a powerful new platform for implementing many of the SI algorithms that our team is developing at Brookhaven Lab,” said Shinjae Yoo, PhD, head of the CDS AI division and principal investigator on the project.
Plasma instabilities
Additionally, a third experiment conducted by the team involved simulated plasma data generated by TORAX. The system is a differentiable tokamak nuclear transport simulator developed by Google DeepMind’s fusion team.
During the experiment, the chip analyzed signals to predict the occurrence of sawtooth instabilities associated with rapid changes in plasma conditions. The results showed that superconducting neural networks could potentially support real-time plasma monitoring and stabilization.
Steven Cowley, PPPL director, emphasized the importance of creating faster control systems for fusion reactors. “We are building the SI-era fusion machines,” he added. “This chip has been taking data from millions of samples directly from a tokamak every second and keeping up.”
Great Sky said the device is the largest superconducting neural chip to date. Similar chips could reportedly find use in quantum computing, cybersecurity and high-speed X-ray data analysis.
The team is currently developing a new chip with a neural network that is ten times larger than the current design. They plan to unveil the hardware at BNL in early 2027 before testing it in fusion experiments on the DIII-D tokamak later this year.
“This is just the beginning,” concluded Jeff Shainline, founder and CEO of Great Sky, in a press release. “This approach to AI hardware has so much to offer in terms of fusion as well as other applications at the forefront of technology.”
