Researchers Developed New Photonic Computing Architecture

A novel system using microring resonators has achieved a computing throughput of 120.8 TOPS.

Updated on Sept. 24, 2026 in Quantum Computing

Researchers Developed New Photonic Computing Architecture

Live Poll

Do you believe breakthroughs in hardware efficiency will meaningfully speed up the progress of AI?

Researchers have developed a photonic computing architecture that utilizes microring resonators as programmable temporal convolution kernels. This advancement enables a multi-channel convolution engine to reach a computing throughput of 120.8 TOPS.

Why it matters

The rapid growth of artificial intelligence creates a demand for higher performance hardware accelerators. This new architecture addresses that need by leveraging the impulse responses of microring resonators to significantly increase efficiency.

The system achieves a compute density of 48.05 TOPS per square millimeter with a single resonator operating at 128 Gbaud. This represents a 160-fold improvement over conventional scalar weighting methods.

The details

By integrating temporal convolution with wavelength- and space-division multiplexing, the design overcomes traditional resonance linewidth constraints. The engine utilizes microring resonators to process data at 5.12 trillion operations per second per unit.

Timeline

  1. September 24, 2026: Research article published on nature.com.

The Tech Race

This innovation challenges the standard reliance on electronic-limited scalar weight architectures in hardware acceleration. It positions photonic systems as a viable path for the next generation of high-speed AI processing hardware.

This breakthrough could eventually lead to faster and more efficient AI tools for consumers and developers. By increasing computing density, future hardware may offer significantly improved performance in smaller, more energy-efficient devices.

The takeaway

This development highlights a major shift in how optical components can be reconfigured for complex computational tasks. Future applications may move beyond academic research as hardware engineers work to integrate these kernels into standard AI accelerator stacks.

Further reading

Learn more about the latest developments in hardware evolution on our Quantum Computing page.

Source note: This article includes information reported by Nature.

Live Poll

Do you believe breakthroughs in hardware efficiency will meaningfully speed up the progress of AI?