tech
In April 2026, University of Cambridge researchers unveiled a new chip material based on a component that mimics how brain neurons connect, storing and processing data in the same place instead of shuttling it back and forth.
The Cambridge team built a memristor from hafnium oxide, a component whose resistance changes based on past current, that stores and processes data in the same place instead of shuttling it between separate memory and processor chips. This 'compute-in-memory' design mimics how brain neurons handle information, and cutting out constant data movement can slash AI energy use by roughly 70 percent.
The lower percentages undersell the scale of savings this approach delivers, since shuttling data across a motherboard is a major, not minor, drain on energy in conventional chip designs.
This falls under neuromorphic computing, research trying to make hardware behave more like a biological brain, which performs enormous computation on only a fraction of a data center's power.
tech
What is the most important reason for taking this approach?What is this numeric identifier called?What is its structural cause?What is this Microsoft chip called?What is this model called?Which was invented first — the fax machine or the telephone?Who is widely regarded as the world's first computer programmer?What does the "http" at the start of a web address stand for?What does "Wi-Fi" actually stand for?What was that first website about?How many satellites does a GPS device need to pinpoint your location?What was the very first message sent over the early internet in 1969?Quration — Quration Play