Computing at the Speed of Light: Australian Photonic Chip Promises a Revolution in AI Energy Efficiency

Researchers from the University of Sydney have presented a prototype nanophotonic chip capable of performing AI computations at the speed of light. This development could become a key solution to the global problem of high energy consumption in data centers, as it uses photons instead of electrons to process data.

How the Chip Completes Tasks in Less Than a Second

Traditional silicon processors generate significant heat during operation due to the electrical resistance of conductors. The nanophotonic chip uses light that passes through structures without resistance or heating.

The photonic chip is extremely small, measuring only a few dozen micrometers in width (approximately the thickness of a human hair). The chip works by forming neural connections within its nanostructures, much like the human brain. As a result, computations take only a few picoseconds, which is the time it takes for light to pass through the device.

This operating principle highlights the possibility of creating AI accelerators that do not require massive amounts of energy or water for cooling. Such technologies make environmentally neutral infrastructure a realistic prospect in the near future.

Other Applications

Scientists have already tested the technology in medicine. The photonic chip is capable of classifying more than 10,000 complex medical images, including MRI scans of internal organs. The classification accuracy ranged from 90% to 99%.