Researchers in Japan have developed a new electronic device that could significantly reshape the future of computing by enabling processors to operate up to 1,000 times faster without generating the excessive heat typically associated with high-performance chips, tells Live Science. The breakthrough centers on a component known as a non-volatile switching element, which addresses one of the biggest limitations in modern computing: the trade-off between processing speed and energy consumption.
Conventional processors rely on electrical currents to switch and store data, a process that generates substantial heat as speeds increase. This heat has become a growing challenge for data centers, artificial intelligence systems, and high-performance computing facilities, all of which require large amounts of electricity for both computation and cooling. The new device uses ultrathin layers of tantalum and an antiferromagnetic material called manganese-tin (Mn₃Sn), allowing data to be switched through magnetic states rather than heat-intensive electrical methods. Researchers demonstrated switching times of approximately 40 picoseconds, far faster than the nanosecond-scale performance of many current memory technologies.
A key advantage of the technology is its non-volatile nature, meaning it can retain information without a continuous power supply. This reduces energy consumption while minimizing waste heat. Laboratory testing also showed strong durability, with the device maintaining stability through more than 100 billion switching cycles. Researchers believe performance could improve further as the components are miniaturized.
If successfully commercialized, the technology could substantially lower the energy requirements of data centers, which are facing rising electricity demands due to the rapid growth of AI workloads. The researchers suggest future computing systems could deliver dramatically higher performance while requiring far less cooling infrastructure. However, significant challenges remain, including large-scale manufacturing, material availability, and integration into existing semiconductor production processes. Prototype systems are not expected before around 2030, but the research offers a glimpse of a future in which faster computing does not come at the cost of increased energy consumption.