Story perspectives

Revolutionary Light-Powered Chip Cuts AI Energy Use 100x

9/11/2025

Summary

Story summary

Researchers at the University of Florida created a light-powered chip that reduces energy consumption for AI tasks by up to 100 times.
The chip improves convolution operations, which contribute to 90% of AI's power use, enabling faster processing with lower energy needs.
This technology could enhance data center efficiency while supporting high-performance AI applications.
The chip achieves 98% accuracy in handwriting classification tests with minimal energy consumption.
High demand is anticipated, prompting significant investment from AI companies despite possible delays.

History

Historical Context

IBM's Watson supercomputer, unveiled in 2011, marked a significant advancement in artificial intelligence by defeating human champions in Jeopardy! This event underscored the urgent need for energy-efficient AI solutions, as Watson's high power consumption became apparent. The University of Florida's new chip, which reduces energy use for AI tasks by up to 100 times, addresses this challenge amid the expanding applications of AI technology.

1 of 2

Story summary

Researchers at the University of Florida created a light-powered chip that reduces energy consumption for AI tasks by up to 100 times.
The chip improves convolution operations, which contribute to 90% of AI's power use, enabling faster processing with lower energy needs.
This technology could enhance data center efficiency while supporting high-performance AI applications.
The chip achieves 98% accuracy in handwriting classification tests with minimal energy consumption.
High demand is anticipated, prompting significant investment from AI companies despite possible delays.

1 / 2