A groundbreaking milestone has just been reached in the field of optical computing. Researchers from the Shanghai Institute of Optics and Fine Mechanics (SIOM), Chinese Academy of Sciences, have successfully developed the world’s first ultra-parallel optical computing integrated chip, named “Meteor One.” For the first time, scientists have achieved over 100-fold parallelism in on-chip photonic computing, setting a new global benchmark in computational performance.

This breakthrough chip demonstrates theoretical peak performance of 2,560 TOPS at an optical clock frequency of 50 GHz—comparable to some of the most advanced GPU chips from NVIDIA. More importantly, it provides a new technical pathway to scale computing power by leveraging the intrinsic advantages of photonics: ultra-high bandwidth, massive parallelism, and low latency.

Dr. Peng Xie, who led the project, explained to DeepTech:

“We achieved over 100-wavelength multiplexed information interaction and computation directly on a photonic chip. This demonstrates the feasibility of high-density, on-chip parallel processing and opens a new direction for scaling optical computing power.”

Pushing Beyond the Boundaries of Electronic Chips

For decades, Moore’s Law has fueled advances in computing, but silicon-based electronics are approaching their physical limits. Issues such as heat dissipation, quantum effects, and soaring power consumption are becoming major bottlenecks, especially in supporting the exponential growth of artificial intelligence workloads.

Photonics offers a compelling alternative. Unlike electrons, photons as bosons can travel at light speed, carry vast amounts of information simultaneously, and operate with far lower latency and power. By extending parallelism through wavelength-division multiplexing, the Meteor One chip achieves a two-order-of-magnitude increase in computational throughput, signaling a fundamental shift in how future high-performance computing could be realized.

A New Architecture for Ultra-Parallel Optical Computing

The research team tackled the long-standing “parallelism bottleneck” by designing an innovative multi-wavelength photonic computing architecture. They integrated multiple critical subsystems—including a self-developed soliton microcomb optical source (>80 nm spectrum, >200 channels), reconfigurable broadband optical computing chips, and high-precision optical matrix driver arrays with more than 256 scalable channels.

This system-level integration enabled the world’s first prototype demonstration of >100-channel wavelength-division multiplexed photonic interaction and computation. Compared with conventional single-wavelength optical computing, this approach achieves two orders of magnitude higher throughput at the same chip frequency and matrix size.

Dr. Xilin Han illustrated the concept with an analogy:

“It’s like upgrading a single-lane highway into a superhighway that can carry a hundred vehicles side by side, massively boosting the traffic flow without changing the road length.”

Transformative Applications Across AI, Data Centers, and Beyond

The implications of this breakthrough are far-reaching. Ultra-parallel photonic computing can dramatically accelerate workloads that rely heavily on large-scale matrix multiplications—such as deep neural networks, embodied intelligence, physical simulations, and real-time image processing.

Its ultra-low latency also makes it ideal for time-sensitive applications with smaller data volumes, including communication switching, edge AI, and autonomous drone coordination. As demand for energy-efficient, high-throughput computing continues to surge, photonic chips like Meteor One could become indispensable for next-generation data centers and AI infrastructure.

Overcoming the “Three Walls” of Optical Computing

To unlock the full potential of photonic computing, researchers must break through three critical barriers:

1. Matrix Scale Wall – Expanding matrix dimensions of photonic chips to boost computational density.
2. Optical Frequency Wall – Pushing higher clock frequencies while maintaining device performance.
3. Parallelism Wall – Exploiting multi-dimensional multiplexing to maximize throughput.

While global players such as TSMC and Caltech have achieved breakthroughs in chip matrix size (512×512) and optical frequency (>100 GHz), respectively, the SIOM team’s innovation lies in solving the parallelism bottleneck.This represents the most practical path to near-term performance scaling, potentially enabling optical computing power to leap ahead by 2–3 orders of magnitude.

A Young, Driven Team Leading the Frontier

This achievement is credited to SIOM’s Photonics and Optical Computing Group, led by Dr. Peng Xie, who previously conducted research at MIT, Oxford, and Nanyang Technological University before returning to China. The team, with an average age of just 28, brings together expertise in optics, electronics, mathematics, and AI to drive rapid innovation.

Notably, every core component of the system—including light sources, interaction chips, computing chips, driver boards, and algorithms—was developed in-house. This end-to-end autonomy not only ensures full technical controllability but also highlights the organizational strength of CAS’s coordinated research model.

Looking Ahead: Toward a New Era of Photonic Supercomputing

The Meteor One prototype marks a crucial first step toward building a complete photonic computing ecosystem. The team is now focused on translating photonic advantages into real-world computational gains by continuously scaling matrix size, frequency, and parallelism.

In the long term, researchers aim to integrate optical computing with neural network architectures, paving the way for photonic supercomputers capable of handling AI, big data, and complex scientific simulations at unprecedented efficiency. With theoretical potential exceeding 5,000 POPS (equivalent to the combined power of 1,000 top-tier GPUs), photonic computing may soon emerge as a disruptive alternative to electronic chips in the post-Moore’s Law era.

As global demand for computing skyrockets, the debut of Meteor One signals not just a technological breakthrough, but the dawn of a new paradigm—a future where light, not electrons, becomes the engine of intelligence.