In order to further reduce the power consumption of next-generation AI high-speed data links and expand the bandwidth of optical channels, with the support of the National Science Council's "Development of 400/800 Gbps SR-8 Optical Transceiver Modules for Next-Generation Data Communications" project, the research team of Professor Jin-Wei Shi of the Department of Electrical Engineering at National Central University has successfully developed a novel single-mode, ultra-high speed, and low-energy consumption vertical cavity surface-emitting laser (VCESL). This achievement not only breaks through the speed and output power limitations of traditional VCSEL light sources, but also, after combining the laser element with single-mode optical fiber, it can achieve a world record of direct modulation bandwidth (40 GHz) at 850 nm wavelength and a distance × data rate product of 1 km×56 Gbps without using any signal processing ICs.
The popularization of optical fiber connection network construction and data centers has brought a new wave of revolution to human civilization. Starting from 2023, the blossoming of artificial intelligence (AI) and advanced integrated circuit (IC) technology has brought about the birth of ChatGPT. This also makes it crucial to move huge amounts of data between graphics processing units (GPUs) and their peripheral switch chips and different data centers at high speed. Currently, the data flow required by an NVIDIA AI cabinet has exceeded the data flow of the entire Ethernet network in the world. Therefore, laser light sources with low power consumption, ultra-high speed, and high-density co-packaging with IC chips are an important trend in today's development. After combining with optical fiber, its laser light source can replace bulky and high-loss RF cables, and can significantly reduce the power consumption in AI cabinets. Therefore, how to effectively expand the bandwidth of the entire optical channel of high-speed laser plus optical fiber has become an important issue to further reduce the power consumption of the next generation of AI high-speed data links.
The core technology of the demonstrated novel single-mode VCSEL is Professor Shi's unique zinc-diffusion structure patent. It is different from the traditional single-mode VCSEL in that it cleverly uses a novel zinc diffusion light-emitting aperture shape. This geometric structure can effectively shorten the carrier transport time and greatly improve the output power and operating speed of the single-mode VCSEL. At present, the zinc-diffusion VCSEL structure has been in mass production by domestic VCSEL chip manufacturers, and its main application is the light source of the proximity sensing module of smartphones. The technology we are currently developing will also be developed and produced in cooperation with these domestic manufacturers. The application of this high-speed, low-energy laser light source in the AI optical link market will be just around the corner. If this achievement can eventually be transformed into mass production of products, it will enable domestic manufacturers to master the core advanced technology of AI data links, making the domestic AI industry chain more complete and conducive to international competition. The research results have been published at the OFC International Conference in 2024, and the complete technical content of the component was also published in the internationally renowned academic journal "IEEE Journal of Selected Topics in Quantum Electronics" in 2025.
Media Contact：
Dr. Yi-Chiung Chao
Program Manager/Assistant Research Fellow
Department of Engineering and Technologies
National Science and Technology Council
Tel: +886 (2)27377941
