Scientists propose ultrafast magnetic-field pulses to cut AI data center energy use by 100x

Researchers at the University of Edinburgh in Scotland have outlined a theoretical approach that could reduce the energy needed for magnetic memory and storage.

In a paper published in Advanced Materials, the team described using ultrafast magnetic-field pulses instead of standard energy-switching techniques. The paper says the approach could cut energy use by up to two orders of magnitude—around 100x. Current methods for switching magnetic memory states can be energy-intensive, a concern as AI data centers increase demand for computing power.

The authors say the framework could also be adapted to electrical currents and ultrafast laser pulses, including technologies being explored for future data storage. Study writer Dr. Elton J.G. Santos said the ideas could have applications beyond the systems examined in the paper.

The work remains theoretical. The researchers proposed practical steps for building prototypes and conducting experiments, but said more work is needed before the approach can be put into practice. The findings suggest a possible route to lower-energy data centers, although they do not establish that the system is ready for deployment.