TLDR: Chinese scientists have developed ‘Meteor-1,’ the world’s first highly parallel optical computing integrated chip. This breakthrough uses light instead of electricity for data processing, achieving a theoretical peak performance of 2,560 TOPS at 50GHz optical frequency. Developed by the Shanghai Institute of Optics and Fine Mechanics and Nanyang Technological University, Meteor-1 aims to accelerate AI and data centers, offering a significant alternative to traditional electronic chips like Nvidia’s GPUs, especially amidst US semiconductor export restrictions.
In a groundbreaking development poised to reshape the landscape of advanced computing, Chinese scientists have unveiled ‘Meteor-1,’ the world’s first highly parallel optical computing integrated chip. This revolutionary innovation, spearheaded by researchers from the Shanghai Institute of Optics and Fine Mechanics (SIOM) and Nanyang Technological University (NTU) in Singapore, marks a significant milestone in China’s pursuit of technological self-reliance amidst escalating trade tensions and stringent US export controls on high-performance semiconductors.
‘Meteor-1’ harnesses the power of light, or photons, for data processing, moving beyond the limitations of traditional electronic chips that rely on electrons. The chip boasts an impressive theoretical peak computing power of 2,560 TOPS (tera-operations per second) at a 50GHz optical frequency. This performance is notably comparable to, and in some aspects, surpasses current high-end GPUs. For context, while Nvidia’s previous flagship RTX 4090 reached 1,321 TOPS, ‘Meteor-1’ offers a substantial increase in computational capacity.
The chip’s advanced architecture, detailed in the research paper ‘Parallel Optical Computing Capable of 100-Wavelength Multiplexing’ published in eLight in 2025, is a key to its capabilities. At its core, ‘Meteor-1’ utilizes a soliton microcomb to generate over 100 coherent light channels and a 5×5 Mach-Zehnder Interferometer (MZI) mesh for programmable matrix computations. Unlike conventional chips that use transistors for parallel processing, ‘Meteor-1’ encodes operations across 100 different wavelengths of light, with each wavelength carrying a slice of the workload in parallel. This multi-wavelength design, coupled with an integrated frequency comb, enables the chip to process over 100 tasks simultaneously without increasing its physical size, effectively multiplying its computational capacity by a hundredfold. The integrated system features a fully self-developed architecture, including a light source chip with an integrated micro-cavity optical frequency comb that spans over 80 nanometers of spectrum and supports more than 200 wavelengths. This innovation eliminates the need for hundreds of discrete lasers, drastically reducing system complexity, power consumption, and cost.
Optical computing offers inherent advantages over electronic computing, including ultra-high speed, broad bandwidth, low power consumption, and minimal latency. These characteristics are crucial for overcoming the fundamental physical limits faced by electronic chips, such as heat build-up, quantum effects, and unsustainable power consumption. The significantly lower power consumption of optical systems, primarily from pump lasers, also paves the way for low-latency, low-power computing at the edge, potentially enabling real-time AI applications on devices like satellites, drones, autonomous vehicles, and even smartphones, where traditional GPUs are often impractical or unsustainable.
The development of ‘Meteor-1’ is a critical advancement for the artificial intelligence industry and data centers, which are currently struggling with soaring computational demands for AI model development, training, and deployment, particularly for generative AI. Experts believe that optical computing could soon outperform traditional electronic processors in speed, energy efficiency, and latency, making it exceptionally promising for next-generation AI, real-time big data analysis, and advanced scientific modeling.
Also Read:
- Nvidia’s AI Supremacy Tested as Major Clients Pursue ‘Sovereign AI’ and Intensified Competition
- KAIST Unveils NPU Technology Dramatically Boosting Generative AI Performance by Over 60%
Beyond its technical merits, ‘Meteor-1’ carries significant geopolitical weight. It is widely seen as a direct response to increasing US export restrictions on advanced semiconductors and AI chips, which have limited China’s access to high-performance GPUs from companies like Nvidia. This breakthrough underscores China’s strategic intent to build sovereign hardware capabilities and achieve technological self-sufficiency, effectively redefining the future of computation. While the hardware is undeniably impressive, experts note that ‘Meteor-1’ is not yet ready for widespread real-life application and commercial production. Nevertheless, its unveiling marks a pivotal moment in the global race for technological supremacy, setting the stage for a new era of innovation and competition in the field of computing.
