Cambridge and USC researchers develop memristors to cut AI energy use by 70%

8 articles · Updated · IO+ · Apr 29

The Cambridge team, led by Dr. Babak Bakhit, engineered a memristor using hafnium oxide infused with strontium and titanium, achieving a million-fold reduction in switching current.
This neuromorphic device integrates memory and processing, mimicking human neurons, and enables electronics to operate efficiently even in extreme environments.
The breakthrough addresses AI’s energy crisis and supports strategic autonomy for Europe and the UK by reducing reliance on energy-intensive legacy computing architectures.

With a 700°C fabrication hurdle, when will these revolutionary AI chips realistically reach the mass market?

How will this 70% energy cut impact the global tech race and Europe's bid for AI autonomy?

As neuromorphic chips advance, how do we govern AI that increasingly thinks more like a human brain?

Could alternative solutions like photonic computing make this new chip obsolete before it even scales?

Will this chip's efficiency reduce the environmental harm from data centers, like heat islands and water use?