New computer chip material inspired by the human brain could slash AI energy use (www.cam.ac.uk)

Researchers from the University of Cambridge have developed a groundbreaking nanoelectronic device using a new type of hafnium oxide that mimics human brain functionality to drastically reduce energy consumption in AI hardware. This memristor design allows information to be both stored and processed locally, potentially slashing energy use by up to 70% compared to traditional computing systems that require significant data movement. The findings, published in *Science Advances*, highlight a shift towards neuromorphic computing, which promises to enhance adaptability and efficiency as AI technologies scale globally. The innovation relies on a two-step fabrication method that forms electronic gates at the interface of hafnium oxide layers by incorporating strontium and titanium. This counteracts the unpredictable behavior seen in conventional memristors, resulting in devices with a million times lower switching currents and hundreds of stable conductance levels. Though currently faced with high fabrication temperatures that challenge integration into chip technology, the researchers are actively working on solutions. If successful, this could lead to chips with significantly reduced energy demands and improved learning capabilities, marking a pivotal advancement in AI hardware development.