As the race for more powerful artificial intelligence (AI) tools accelerates, China is placing major bets on optical computer chips to overcome the speed and energy limits of traditional electronic hardware. These semiconductor chips, which process data using light instead of electricity, are at the center of China’s strategy to lead the next phase of AI development.
The push comes as China faces increasing restrictions from the United States on access to advanced electronic chips and the equipment needed to produce them. These components are critical for training and running large-scale AI models. In response, China is turning toward optical computing as an alternative path to high-performance systems.
Zengguang Cheng, a materials scientist at Fudan University in Shanghai, said the country has strong motivation to explore new approaches in computing. He noted that China’s 14th Five-Year Plan includes goals related to photonics and quantum technologies, backed by steady government funding.
China shifts focus to optical computer chips
Optical chips, also known as photonic chips, have several advantages over their electronic counterparts. They use photons to transmit information, allowing for faster performance with less heat and lower energy consumption. While these chips are already used in areas such as sensors and medical devices, using them for complex AI tasks remains a technical hurdle.
China is ramping up investment in optical computer chips as it looks to power the next wave of AI with faster, energy‑efficient technology.#China #OpticalComputing #AI #Semiconductors #PhotonicChips #TechNews pic.twitter.com/ODKK7h0WRR
— Tom Marvolo Riddle (@tom_riddle2025) January 30, 2026
Photonic chips operate by adjusting light properties such as amplitude and phase, rather than using transistors to control voltage. That makes them energy-efficient but more difficult to scale and train, according to Yitong Chen, the lead author of the study and an electronic engineer at Shanghai Jiao Tong University. Until recently, most optical chips could only handle limited functions like image classification.
Chen and his team reported a breakthrough last month with the development of LightGen, the first all-optical chip designed to handle advanced generative AI models. Built using layers of engineered metasurfaces that manipulate light at the nanoscale, the chip contains millions of photonic neurons.
LightGen pushes the boundaries of AI with photonic processing
The team also designed a custom training algorithm for optical systems. They said LightGen can generate images, edit videos, and create 3D scenes at greater speed and efficiency than high-end processors such as NVIDIA’s A100.
Ben Eggleton, a physicist at the University of Sydney, described the chip as a strong example of how optical systems can complete specific AI tasks while using far less energy.
Over the past five years, research into optical chips has surged, with China leading global output. In 2025, Chinese researchers published 476 papers on the subject, more than any other country, according to data from the Dimensions database analyzed by Nature. The figure marks a tenfold increase since 2017.
