Artificial intelligence is everywhere lately, but it requires a massive amount of electricity. Most of our current computer chips waste energy because they have to constantly shuffle data back and forth between memory and processing units. It’s like having to walk to a different room every time you need to remember a phone number. To fix this, researchers at the University of Cambridge have developed a new device that works more like a human brain.

This new hardware uses a “memristor,” a component that mimics how neurons connect. By storing and processing information in the same spot, these devices could cut energy use by as much as 70%. It’s a step toward “neuromorphic” computing, systems that don’t just store data bits, but learn and adapt like we do.

“Energy consumption is one of the key challenges in current AI hardware,” said lead author Dr. Babak Bakhit. “To address that, you need devices with extremely low currents, excellent stability, outstanding uniformity across switching cycles and devices, and the ability to switch between many distinct states.”

How the Computer Chip Works

Most older memristors rely on tiny filaments that break and regrow, which makes them hard to control. The Cambridge team tried something different and created a thin film made of hafnium oxide mixed with strontium and titanium. Instead of messy filaments, this film uses tiny electronic gates to change its resistance smoothly.

In the lab, these devices survived tens of thousands of cycles and used currents a million times lower than traditional versions.

“Filamentary devices suffer from random behavior,” Bakhit said. “But because our devices switch at the interface, they show outstanding uniformity from cycle to cycle and from device to device.”

The Road Ahead

Bakhit spent nearly three years facing failures before a late-night tweak to the manufacturing process finally worked. However, the process currently requires temperatures around 700°C, which is too hot for standard factory lines.

“This is currently the main challenge in our device fabrication process,” Bakhit said. “But we’re now working on ways to bring the temperature down to make it more compatible with standard industry processes.”