China just broke the world record for fiber transmission, and the AI data center race will never be the same | Daily Reports Online
- Air-filled fiber carried 51.3 Tb/s across 206.5 km unaided
- Researchers push 1.2 Tb/s through each wavelength across record distances
- AI infrastructure increasingly depends on networks moving data faster everywhere
Chinese fiber manufacturer YOFC says it has completed a hollow-core fiber transmission trial that reached 51.3 Tb/s over 206.5 km without regeneration.
The demonstration involved collaboration with China Telecom and optical equipment manufacturer Dekoli using a live network rather than laboratory conditions alone.
Researchers achieved a transmission rate of 1.2 Tb/s per wavelength while avoiding intermediate signal regeneration equipment along the entire route.
A different kind of fiber built around air instead of glass
Unlike conventional optical cables that guide light through solid silica cores, hollow-core fiber transmits signals through air-filled channels instead.
That architectural difference allows light to travel faster while reducing several optical distortions that traditionally limit transmission efficiency over distance.
YOFC previously stated that its hollow core technology could deliver 31% lower latency and transmission speeds improving by 47%.
The company now claims the latest trial establishes the highest unrepeatered wavelength division multiplexing capacity ever demonstrated under field conditions.
Researchers described the achievement as the world’s first field deployment combining 1.2 Tb/s wavelengths with a 206.5 km span.
Previous demonstrations managed comparable transmission rates over roughly 20 km, while longer experiments generally sacrificed substantial amounts of overall capacity.
The trial relied exclusively on erbium-doped fiber amplifiers instead of remote-pumped amplification systems often required for comparable distances.
Commercial hollow core deployments have historically struggled with signal attenuation, making long unrepeatered transmission distances difficult to sustain economically.
Network cables become the next computing bottleneck
The research team tackled those limitations by using adaptive allocation techniques that independently adjusted channel rates and optical power across wavelengths.
That approach allowed hybrid transmission settings while reducing losses associated with gas absorption effects unique to air-guided optical signals.
Engineers also developed a high-power amplifier capable of producing 33.5 dBm output while maintaining relatively uniform gain characteristics.
Because transmitting optical power approaching 2.24 W introduces operational risks, several automatic protection systems monitored link behaviour continuously throughout testing.
Safeguards included anomaly detection systems, automated shutdown functions and alarm-triggered responses designed to prevent expensive equipment failures during operation.
The timing of the experiment coincides with accelerating demand from AI tools requiring unprecedented movement of information between data centers worldwide.
Large GPU clusters increasingly depend on networking performance, creating constraints that processors alone cannot solve through additional computing resources.
Lower-latency transmission could allow operators to distribute facilities farther apart without incurring penalties that affect training and inference performance.
YOFC believes the trial marks progress toward wider deployment, although competing hollow core ecosystems are rapidly emerging outside China’s supply chains.
Whether such experiments eliminate bandwidth bottlenecks remains uncertain, although networking limitations increasingly appear as important as computing limitations themselves.
Via Tom’s Hardware
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds. Make sure to click the Follow button!
And of course you can also follow TechRadar on TikTok for news, reviews, unboxings in video form, and get regular updates from us on WhatsApp too.
