Accelerating Chipmaking Innovation for the Energy-Efficient AI Era

# AI Chips Need a New Approach to Avoid Wasting Energy The race to build faster AI systems is hitting a surprising bottleneck: moving data around uses as much electricity as actually processing it, so companies need to rethink how they design chips from the ground up rather than improving individual pieces in isolation. This means coordinating improvements across three areas—how transistors switch power on and off, how quickly data reaches processors from storage, and how chips are physically stacked together—all at the same time instead of separately.