Scientists find a way to slash computer memory energy use by orders of magnitude

Internet searches, AI-generated images, recommendation systems, scientific simulations, and large language models all depend on enormous amounts of information being created, moved, stored, and analyzed. As AI becomes more deeply integrated into everyday life, industry, and science, the global need for computing power and data storage continues to climb.

AI's Growing Energy Demand

That expansion also brings a major challenge: electricity use. Data centers already require huge amounts of power, and their energy demands are expected to rise substantially in the coming decades. Without significant improvements in efficiency, ICTs could eventually represent a sizable share of worldwide electricity consumption and carbon emissions.

Finding ways to make computing more energy efficient is therefore becoming increasingly important as demand for digital services accelerates.

Researchers at the University of Edinburgh have developed a new theoretical framework that could help reduce the amount of energy needed to store and manipulate digital information (bits, represented as "0"s and "1"s) in future magnetic memory technologies.

A More Efficient Way to Switch Magnetic Memory

At the heart of magnetic memory is the ability to switch magnetic states, which allows digital information to be changed and controlled.

Instead of using conventional methods for designing magnetic switching processes (which is the basic mechanism behind data manipulation), the researchers turned to Optimal Control Theory, a mathematical approach used to determine the most efficient way to reach a specific goal.

Using this method, the team created a framework for designing ultrafast magnetic-field pulses that can switch magnetic states while consuming as little energy as possible. The calculations also take realistic experimental limitations into account, making the approach more relevant to potential future devices.

Moving Closer to a Fundamental Energy Limit

Computer simulations suggest that the method could lower switching energy by several orders of magnitude compared with leading memory technologies used or being developed today, including DRAM, STT-MRAM and emerging SOT-MRAM devices.

Even more strikingly, the predicted energy requirements move future magnetic memory much closer to the Landauer limit (the fundamental thermodynamic limit) defining the minimum amount of energy required to process a single bit of information.

That limit represents a fundamental boundary imposed by physics, so approaching it would mark a major advance in the effort to make computing as energy efficient as possible.

The framework, described in Advanced Materials, also goes beyond theoretical calculations. It includes practical guidance for possible implementation, including optimized device designs and methods for delivering magnetic fields. These recommendations could help researchers eventually test the concept experimentally.

Potential Beyond Magnetic Fields

Dr. Elton Santos from the Institute for Condensed Matter Physics and Complex Systems, University of Edinburgh, who led the research, said:

"Every digital operation has an energy cost, and that cost becomes increasingly important as AI and data-intensive technologies continue to expand. Our work shows that, by carefully designing how a magnetic field changes in time, magnetization can be switched far more efficiently than with conventional approaches."

He continued:

"Although we first developed the theory using magnetic field pulses, the mathematics is far more versatile than that. The same framework can be adapted to electrical currents and even ultrafast laser pulses, which are among the most cutting-edge technologies for future data storage. That means the ideas developed here could have applications far beyond the systems we studied. It seems that we may have just found the next best thing."