Every time your device stores or retrieves data, it burns energy flipping magnetic bits. Multiply that by a trillion operations per second, then multiply again by every AI training run and every data center humming around the clock — the electricity demand adds up fast. Researchers at the University of Edinburgh think they’ve found a smarter path forward, using mathematics to reshape how those bits get switched.
Published July 14, 2026 in Advanced Materials, the paper by Mohammad H. Badarneh, PeiYu Cai, and Elton J. G. Santos describes a framework built on optimal control theory — essentially, using math to design precisely shaped magnetic-field pulses rather than the fixed, blunt-force pulses current memory technologies rely on. The goal is to approach the Landauer limit, the thermodynamic floor for how little energy physically needs to be spent to process a single bit of information.
In simulations, the results are striking:
- Switching energy reduced by several orders of magnitude compared with DRAM, STT-MRAM, and SOT-MRAM
- Materials tested include van der Waals magnets — layered magnetic crystals like Fe3GaTe2, Fe3GeTe2, and CrSBr — switching on picosecond timescales
- The paper includes guidance on optimized device designs and field-delivery methods, pointing toward experimental testing rather than purely abstract modeling
“Every digital operation has an energy cost, and that cost becomes increasingly important as AI and data-intensive technologies continue to expand.” — Elton J. G. Santos, University of Edinburgh
Promising Simulation, Unproven Silicon
The gap between a compelling theory and a manufacturable memory chip is where promising ideas routinely meet physical reality.
Here’s the honest version: this is a theoretical framework, not a product. The distance between a compelling simulation and a shipping chip is vast. You should be interested, not excited. Yet.
That said, the framework’s adaptability matters. Santos notes it could be extended to electrical currents and ultrafast laser pulses — meaning the math isn’t locked to one narrow use case. According to the International Energy Agency, data centers already account for roughly 1–2% of global electricity demand, a figure climbing alongside AI adoption. If memory switching could approach the Landauer limit at scale, efficiency gains would ripple across the entire computing stack.
The next step is getting this out of simulation and into actual devices. If experimental validation follows, it could become a design principle across multiple memory and switching technologies — not just a single lab curiosity.
