Researchers at the University of Tokyo developed a magnetic switching device that operates at speeds up to 1,000 times faster than existing AI accelerators, while consuming significantly less energy and generating minimal heat. This innovation aims to tackle overheating and battery drain in electronic devices, which could enhance the efficiency of computers and smartphones.
The research was published in the journal Science earlier this week and builds upon a study published in Nature in January 2025. The new device utilizes a method to flip a binary magnetic state at picosecond speeds, a considerable advancement over the nanosecond-scale switching typical in silicon-based processors. The study addresses heat generation, which increases with processor speed and results in excessive power consumption in data centers.
The researchers constructed a spintronic device using a manganese and tin compound (Mn3Sn), known for its antiferromagnetic properties. This type of device leverages both the charge and spin of electrons, allowing for more efficient data processing, storage, and transmission compared to traditional semiconductors.
In their proof of concept, the team demonstrated that sending a 40-picosecond electrical pulse through the antiferromagnet flips its magnetic state with minimal resistive heat generation. This process consumes less energy than current AI accelerators, raising expectations for the development of more efficient AI hardware.
One picosecond is equivalent to one trillionth of a second, making it 1,000 times shorter than a nanosecond. Should this technology transition from research to commercial use, it could be advantageous for cloud-based quantum services, potentially making optical quantum computing more accessible. Professor Tomo Nakatsuji stated, “there is (also) a possibility that data that takes an hour to download can be processed in one second.”
It is important to note that while increasing the binary state switching speed 1,000 times represents a significant advancement, it does not translate to a thousandfold increase in overall computing speed due to the complexities of computer systems, which depend on multiple hardware and software components working together.
