China's hollow-core fiber trial pushes 51.3 Tb/s over 128 miles without signal regeneration — milestone targets AI-era networking bottlenecks

At a glance:\n- YOFC, China Telecom and Dekoli achieved a world‑record 51.3 Tb/s over 128 mi (206.5 km) using hollow‑core fiber without any repeaters or remote‑pumped amplifiers.\n- The trial delivered 1.2 Tb/s per wavelength via WDM and relied solely on conventional erbium‑doped fiber amplifiers (EDFAs), a first for unrepeatered HCF spans.\n- The breakthrough cuts latency, reduces nonlinear distortion and opens new capacity for AI‑era data‑center and long‑haul networks.\n\n## What happened\nThe trial was announced on June 16 by Chinese firm Yangtze Optical Fiber and Cable Joint Stock Limited Company (YOFC) in collaboration with state‑owned China Telecom and optical equipment maker Dekoli. Using a self‑developed hollow‑core fiber (HCF) wavelength‑division multiplexing (WDM) system, the partners transmitted an aggregate 51.3 Tb/s over roughly 128 miles (206.5 km) without any signal regeneration. The demonstration achieved a new world record for unrepeatered WDM capacity‑distance performance, employing only erbium‑doped fiber amplifier (EDFA) amplification‑a first for live network HCF deployments.\nPreviously, China Telecom had demonstrated 1.2 Tb/s over a single wavelength in July 2024, but only over a 20‑kilometer span. Earlier research had pushed unrepeatered HCF spans past 300‑kilometers, yet at far lower overall capacities. YOFC's achievement combines high per‑channel data rates, long reach, and conventional amplification in a real‑world field trial.\n\n## Why it matters\nHollow‑core fiber's air‑filled core offers several key advantages that align with AI networking demands:\n- Light travels roughly 1.5 times faster through air than through glass, cutting latency.\n- The air core sidesteps much of the nonlinear distortion and dispersion inherent in silica.\n- YOFC claims the technology can deliver 31 % lower latency, 47 % faster transmission speeds, and near‑zero optical nonlinearity compared with conventional solid‑core fiber.\nThese properties give operators the capacity headroom and reduced bottlenecks needed as hyperscalers build ever‑larger GPU clusters. Lower latency also enables facilities to be sited farther from expensive, power‑constrained hubs without sacrificing speed, making HCF attractive for both data‑center interconnects and long‑haul links.\nThe market is already moving quickly: Microsoft's 2022 acquisition of Lumenisity led to manufacturing deals with Corning and Heraeus in September 2025, while AWS touts its own HCF claiming a 30 % latency improvement. Corning supplies fiber to Meta, Lumen, and others, and is expanding a North Carolina facility backed by Nvidia. These efforts suggest a converging market, with YOFC's breakthrough offering a non‑Western reference point for future deployments.\n\n## Technical innovations\nThe collaborators credited two main innovations for the record‑breaking performance:\n- System‑level optimization: A self‑developed scheme adapts per‑wavelength rate and channel power allocation to link conditions, enabling hybrid transmission across multiple data rates, channel spacings, and power levels. This reduces capacity losses caused by gas‑absorption peaks specific to guiding light through air.\n- Amplifier hardware: Researchers built a high‑power amplifier using a cascaded dual‑gain‑unit architecture and a multi‑element doping design, achieving a maximum output of 33.5 dBm (≈2u202fW) while maintaining flat gain across the operating band. The higher‑power, flatter amplification allowed an unrepeatered span without remote‑pumped boosters.\nBecause pushing that much power over a live optical link carries a risk of failure, the system incorporated safeguards such as optical‑path power anomaly detection, automatic interlock shutdown, and alarm‑linked response mechanisms to catch faults before equipment damage.\n\n## Industry context and competition\nEarlier HCF research had demonstrated unrepeatered spans exceeding 300‑kilometers, but at substantially lower overall capacities. China Telecom's July 2024 demonstration of 1.2 Tb/s over just 20‑kilometers highlighted the distance‑capacity trade‑off that YOFC's trial now overcomes.\nWestern players are also racing to commercialize HCF. Microsoft, via its Lumenisity acquisition, is scaling production with Corning and Heraeus, while AWS touts its own HCF claiming a 30 % latency edge. Corning supplies fiber to Meta, Lumen, and others, and is expanding a North Carolina facility backed by Nvidia. These efforts suggest a converging market, with YOFC's breakthrough offering a non‑Western reference point for future deployments.\n\n## Future outlook and deployment challenges\nThe trial proves that commercial hollow‑core fiber can support long‑haul and data‑center interconnects with the capacity and latency benefits required for AI workloads. However, loss remains a trade‑off; ongoing work focuses on further reducing attenuation to make HCF competitive with mature silica fiber.\nLooking ahead, scaling the adaptive optimization scheme and integrating HCF into existing network architectures will be critical. As hyperscalers continue to expand GPU clusters, breakthroughs like this one will be essential to avoid network bottlenecks. The technology's promise—lower latency, higher capacity, and near‑zero nonlinearity—positions hollow‑core fiber as a cornerstone for the next generation of AI‑driven infrastructure.\n