The GIST: Australian researchers at the University of Sydney have developed a groundbreaking ultra-compact artificial intelligence (AI) chip that performs computations at the speed of light using photons instead of electrons.
Source: Inverse-designed nanophotonic neural network accelerator. Credit: Nature Communications (2026). DOI: 10.1038/s41467-026-68648-1
Revolutionary Photonic AI Technology
Researchers at the Sydney Nano Hub, University of Sydney, have successfully built a prototype AI chip that relies on light particles—photons—to perform calculations. Unlike traditional chips that use electrons and generate heat due to electrical resistance, this nanophotonic chip operates without significant energy loss, offering a highly energy-efficient alternative for future computing.
The chip’s nanostructures, each tens of micrometers in size (roughly the width of a human hair), form an artificial neural network capable of performing complex calculations. Computation occurs on a picosecond timescale—trillionths of a second—allowing the AI chip to process data at light speed.
Professor Xiaoke Yi, director of the Photonics Research Group, emphasized the significance of the innovation:
"We've re-imagined how photonics can be used to design new energy-efficient and ultrafast computer processing chips. Artificial intelligence is increasingly constrained by energy consumption. This research demonstrates neural computation using light, enabling faster, more compact AI accelerators."The study, published in Nature Communications, shows that AI models can be directly embedded into nanoscale photonic structures that manipulate light to perform mathematical operations necessary for machine learning.
Validation and Performance
To test the technology, researchers trained the nanophotonic chip to classify over 10,000 biomedical images, including breast, chest, and abdomen MRI scans. Both simulations and experiments showed the chip achieved classification accuracies between 90% and 99%.
This demonstrates that photonic AI chips can deliver high-performance computation while drastically reducing the energy demands associated with conventional data centers, which often require massive amounts of electricity and cooling systems.
Ph.D. student Joel Sved, who contributed significantly to the chip’s design, explained:
"The prototype shows how intelligence can be embedded directly into nanoscale photonic structures, opening pathways for sustainable, high-speed AI hardware."Why Photonics Matters
Photonics, the study of controlling light particles, has long powered technologies like lasers, fiber-optic communications, and medical imaging. Using photonics for computing, however, is a relatively recent development driven by the surge in AI demand.
The University of Sydney’s Photonics Research Group has a decade-long history of advancing photonics in areas such as wireless communications and advanced sensing technologies. This research builds on that foundation to push AI hardware toward higher speeds and energy efficiency.
Following the successful prototype, Professor Yi’s team is now developing larger-scale photonic neural networks that could redefine AI computation infrastructure.
Read More: ‘WhatsApp Plus’ subscription: Everything you need to know about upcoming update
FAQs
How is a photonic AI chip different from traditional computer chips?
Traditional chips use electrons to process information, which generates heat due to electrical resistance. Photonic chips use photons (light particles), which travel without resistance, enabling faster computations with minimal heat generation.
What is the size of the nanostructures on the chip?
The nanostructures are tens of micrometers in size—roughly comparable to the width of a human hair—but are engineered to perform complex AI calculations at the nanoscale.
How fast can the photonic chip perform calculations?
The chip operates on a picosecond timescale (trillionths of a second), effectively performing calculations at the speed of light.
What kind of AI tasks can it handle?
The prototype has been trained to classify over 10,000 biomedical images, including MRI scans, achieving 90–99% accuracy. It can potentially be adapted for other AI applications requiring rapid and energy-efficient computation.
What are the environmental benefits of using photonic AI chips?
By reducing energy consumption and heat generation, photonic chips can lower the energy footprint of data centers and AI infrastructure, supporting sustainable computing at scale.
Conclusion
The University of Sydney’s photonic AI chip represents a major step forward in the convergence of light-based technology and artificial intelligence. By harnessing photons for computation, researchers have demonstrated a pathway toward faster, ultra-compact, and energy-efficient AI hardware.
This technology not only promises to reduce the environmental footprint of future computing systems but also paves the way for next-generation AI infrastructure capable of meeting the growing global demand for intelligent, sustainable computing. As research progresses, larger-scale photonic neural networks could redefine the future of high-performance AI hardware.
