YOFC Unveils New AI Fiber Tech
At MWC Barcelona 2026, Yangtze Optical Fibre and Cable (YOFC) will unveil a new Hollow-Core Fibre (HCF) solution. The technology is designed to provide ultra-low latency optical communication to strengthen the infrastructure required for advanced AI.
Hollow-Core Fibre (HCF) guides light through an air-filled channel instead of a solid glass core. This allows light to travel approximately 30-47% faster, significantly reducing latency compared to conventional single-mode fiber. This fundamental design shift also minimizes signal distortion and allows for the transmission of higher optical power. For AI applications, this ultra-low latency enables distributed computing clusters to function as a single, cohesive system, even when spread across tens of kilometers. This is critical for the intense and time-sensitive workloads of training and inference in advanced AI models. The concept of guiding light through a hollow core dates back decades, but early versions suffered from high signal loss, known as attenuation. Recent breakthroughs, such as the development of Nested Anti-Resonant Nodeless Fiber (NANF), have reduced this loss to levels comparable to or even better than traditional fiber, making it commercially viable. This technology is a key component of the "AI-2030" strategy for Yangtze Optical Fibre and Cable (YOFC), championed by company president Zhuang Dan. The strategy aims to position the company as a global leader in AI-driven optical infrastructure by accelerating the adoption of these advanced technologies. YOFC is not alone in this pursuit. Tech giants like Microsoft are also investing heavily in HCF to optimize their global cloud infrastructure for AI workloads. Microsoft has discussed plans for tens of thousands of kilometers of HCF to improve its Azure network's performance and latency. The extended reach of HCF could reshape data center geography. It enables facilities to be built farther apart and in more diverse locations, potentially closer to renewable energy sources, without the typical performance penalties associated with distance.