Cambridge Researchers Build Brain-Like Chip That Could Cut AI Energy Use by 70%
Researchers at the University of Cambridge have engineered a new nanoelectronic device using modified hafnium oxide that mimics how neurons process and store information. The neuromorphic chip combines memory and processing in a single unit, operating with ultra-low power. Scientists say it could reduce AI energy consumption by up to 70%.
Researchers at the University of Cambridge have built a new nanoelectronic device that mimics how neurons process and store information, potentially cutting AI energy consumption by up to 70%.
The chip uses modified hafnium oxide to create a neuromorphic architecture. Unlike traditional chips that separate memory and processing into distinct units, this device combines both functions in a single component. That design mirrors how biological neurons work.
The chip operates with ultra-low power and shows stable conductance levels and biological learning behaviors in testing. Scientists say it could dramatically reduce the energy demands of running large AI models.
AI data centers currently consume enormous amounts of electricity. The energy cost of training and running large language models has become a major concern for technology companies and policymakers alike.
The Cambridge research was published in April 2026 and reported by ScienceDaily. It represents one of several efforts to build more efficient AI hardware.
Separately, Cognichip raised $60 million to use AI for designing AI chips, aiming to cut development costs by over 75% and timelines in half. Google and Intel are also deepening their AI infrastructure partnership to co-develop custom chips amid a CPU shortage.
Google's latest Tensor Processing Unit chips are being closely watched as an alternative to Nvidia for AI compute. Elon Musk outlined plans for "Terafab," a large-scale AI chip manufacturing project in Austin, Texas, intended to support Tesla, SpaceX, and xAI.
