China's First Energy-Efficient Carbon Nanotube AI Chip

Chinese scientists have developed a novel tensor processing unit (TPU) using carbon nanotubes instead of silicon, enhancing energy efficiency for AI tasks. This new chip, described in Nature Electronics, operates at just 295 microwatts while delivering 1 trillion operations per watt. Built with 3,000 carbon nanotube transistors and a systolic array architecture, it processes data faster and more efficiently than traditional silicon-based chips. The innovation could significantly advance energy-efficient AI technology.