Hollow-core optical fiber is not a new concept, but it was only in recent years that it stepped out of laboratory papers and became an option on operators' bidding lists. Its core proposition is simple yet bold: since the loss, nonlinearity, and latency of glass are approaching their limits, let the light leave the glass and return to the air.
Traditional optical fibers trap light in a silica core, while hollow-core fibers use a microstructured cladding to surround a hollow channel, where light propagates at a speed close to the speed of light in a vacuum. This is not merely a material substitution, but a rewriting of the underlying logic of optical fiber communications.
01. A "Capacity Ceiling" Quietly Approaching
Over the past half-century, the progress of optical communications has almost been built on a default premise: light propagates in glass. Since Corning produced the first low-loss silica optical fiber in 1970, humanity has steadily compressed optical signals more densely and transmitted them further, with wavelength division multiplexing turning a single optical fiber into hundreds of parallel lanes. But this approach is hitting its ceiling.
The communication capacity of single-mode optical fibers has approached the "nonlinear Shannon limit"—no matter how much the transmission power increases, the signal-to-noise ratio no longer improves linearly, and the growth of the system's capacity-distance product encounters a wall that is difficult to surmount. More pressing are the changes on the demand side. AI large model training has turned data centers into "bandwidth-devouring monsters".
Data from research institution CRU shows that by 2025, the demand for optical cables in data center applications will account for about 5% of the total global optical cable demand, with this proportion reaching as high as 14% in North America. Meta (META.O) and Corning (GLW.N) signed a data center optical fiber cable supply agreement worth up to USD 6 billion, and NVIDIA (NVDA.O) has set 2026 as the "first year of commercialization of silicon photonics".
Computing power is surging, but the road that carries it is quickly becoming not wide enough.
02. When Light No Longer Travels Through Glass
The logic of hollow-core optical fiber to break the bottleneck can be summarized in one sentence: let the light return to the air. The latency bottleneck of traditional optical fibers comes from the glass—the propagation speed of light in silica is only about 69.5% of the speed of light in a vacuum. Hollow-core optical fibers hollow out the core, allowing light to travel in an air channel close to a vacuum, with the speed directly surging to 99.7% of the speed of light.
This seemingly simple physical substitution has triggered chain reactions far beyond expectations: transmission latency is reduced by about 31%, nonlinear effects are reduced by more than three orders of magnitude, and signal distortion during long-distance transmission is significantly decreased.
Progress in attenuation data is equally encouraging, and the pace of commercial deployment is faster than expected. In July 2025, China Mobile (600941) launched China's first commercial anti-resonant hollow-core optical fiber line in Guangdong, reducing the two-way transmission latency between the data centers of the Shenzhen and Hong Kong stock exchanges to 1.07 milliseconds. In October of the same year, China Unicom (600050) completed the cross-border commercial interconnection of hollow-core optical fibers between Shenzhen and the Zhiyun Data Center in Tseung Kwan O, Hong Kong, China, reducing network latency by 32%, making it the world's first case to achieve a direct connection of "submarine cable center + data center + financial center" through hollow-core optical fibers. In November of the same year, a 100-kilometer-level co-fiber transmission experiment using hollow-core optical fibers by China Telecom (601728) achieved a transmission rate of 37.6 Tbps.
In October 2026, in Zhongwei, Ningxia, hollow-core optical fibers were again the first to be put into commercial use between intelligent computing centers 30 kilometers apart, becoming a "maglev" track for optical signals in the era of AI computing power. The financial industry was the first to feel the value of this new road. The commercial dedicated line of hollow-core optical fibers by Hengtong Optic-Electric (600487) has been put into financial business operation, with latency reduced by 33% compared to traditional optical fibers. In the world of high-frequency trading where every millisecond counts, this 33% is real money.
03. Turning One Optical Fiber into Several
If hollow-core optical fiber solves the problem of "fast," then multi-core optical fiber solves the problem of "more." The idea is equally intuitive: since the capacity of a single core has reached its limit, pack more cores into the same optical fiber.
Within a cross-section as thin as a hair, multi-core optical fibers can integrate 4, 7, or even more independent signal channels, which is equivalent to directly widening the information highway from a "single lane" to "multiple lanes." Moreover, its outer diameter remains unchanged. In scenarios with extremely limited space such as submarine optical cables, this advantage is almost decisive.
Commercial deployment has already run ahead of expectations.
In 2026, the multi-core optical fiber submarine cable system built by FiberHome (600498) from Guishan Island to Wailingding Island in Zhuhai was officially put into commercial use. Adopting a self-developed 7-core multi-core optical fiber networking, this is the world's first large-scale commercial multi-core optical fiber submarine cable system. Hengtong Optic-Electric, in conjunction with China Mobile, opened the world's first S+C+L tri-band ultra-low loss multi-core optical cable line in Qingdao, Shandong, with the capacity of a single optical fiber reaching more than 5 times that of traditional optical fibers. HMN Tech is pushing further in the direction of the deep sea, achieving ultra-long-distance transmission of 11,250 kilometers based on 4-core weakly coupled optical fibers. Coupled with the mature 32-fiber-pair and 48-fiber-pair submarine cable space division multiplexing solutions, the system can achieve Petabit-level transmission capacity.
From offshore to transoceanic, multi-core optical fibers are redefining the capacity ceiling of submarine optical cables.
04. Standards and Costs: Two Hurdles for Scaling Up
Just because the technology works does not mean the industry works. There are still two hurdles in front of next-generation optical fibers.
The first is standards. The attenuation measurement methods, splicing specifications, and compatibility indicators with traditional optical fibers for hollow-core optical fibers are still under discussion. In July 2026, the ITU-T SG15 plenary meeting agreed to initiate a technical report on hollow-core optical fibers, which is the first research report project specifically for hollow-core optical fibers at the International Telecommunication Union level. The China Communications Standards Association also added hollow-core optical fiber test method standards in February 2026.
As for multi-core optical fibers, ITU-T initiated standards related to weakly coupled multi-core optical fibers in March 2025, which are expected to be released in 2028, and multiple domestic and foreign manufacturers are jointly promoting them. Standards lagging behind technology is not uncommon in the communications industry, but for a technology that is going to enter the national backbone network, the lack of standards means uncontrollable deployment risks.
The second is cost. According to publicly available bidding data from operators, the transaction unit price of domestic hollow-core optical fibers is concentrated at CNY 25,000 to CNY 36,000 per fiber-kilometer, and the price difference with conventional G.652D ordinary single-mode optical fibers (about CNY 20 per fiber-kilometer) is close to 2,000 times. Li Junjie, Deputy Dean of the Research Institute of China Telecom, revealed at the 2026 Optical Expo that the unit price of hollow-core optical fibers has gradually dropped to over CNY 20,000 per fiber-kilometer, but the efficiency of post-processing technology still severely restricts the delivery cycle. Matt Rehder, Vice President of Core Network at AWS, a subsidiary of Amazon (AMZN.O), frankly stated that the real challenge lies in manufacturability, which is directly related to cost—this type of optical fiber is extremely difficult to manufacture, and the current yield rate and manufacturing costs are not ideal.
However, the solution to the cost problem often does not lie in the optical fiber itself. A deployment cost model covering five application scenarios shows that optical fibers and cables account for only 5% to 10% of the outside plant deployment costs, and hollow-core optical fibers can eliminate the need for expensive coherent transceivers by eliminating chromatic dispersion limits, with the cost difference of a single transceiver reaching USD 1,000 to USD 2,000. In other words, when the total system cost is taken into consideration, the economic tipping point of hollow-core optical fibers may come earlier than the unit price suggests.
05. A Race Over "Right of Way"
Global major players have already expressed their judgment on next-generation optical fibers through actions. Ding Qi, an analyst at CITIC Securities, citing data from CRU and LightCounting at the 2026 China Optical Communications High-Quality Development Forum, pointed out that the consumption of optical cables for AI applications grew by 138% in 2024 and continued to grow by 80% in 2025. He estimates that the market space for hollow-core optical fibers will be around the level of CNY 5 billion in 2026, and will gradually enter the CNY 10 billion level by 2027 to 2028.
The forecast from Research and Markets is more macroscopic: the next-generation optical fiber market will grow from USD 13.52 billion in 2025 to USD 39.23 billion in 2032, with a compound annual growth rate of 16.43%.
The layout pace of overseas giants is equally compact. Microsoft (MSFT.O) announced a plan to deploy 15,000 kilometers of hollow-core optical fibers within two years at the end of 2024, and has outsourced the manufacturing process to Corning and Heraeus Covantics. AWS is deploying hollow-core optical fibers to connect about 10 data centers, and its Vice President of Core Network believes that hollow-core optical fibers "may replace traditional optical fibers in the future and become the standard choice for long-haul backbone networks".
On the domestic front, YOFC (601869) launched the INFINITE-2030 plan, aiming to promote more than 50 commercial and pilot projects for hollow-core optical fibers; previously, the company has deployed more than 10 commercial and pilot projects in Asia, Europe, and the Americas.
From the attenuation curves in the laboratory to the commercial submarine cables on the seabed, from the centralized procurement bidding of operators to the standard initiation of ITU-T, next-generation optical fibers are completing the transition from "technical narrative" to "industrial fact." Its ultimate form may not yet be completely clear—the route debate among hollow-core, multi-core, and large-core-count optical cables is still ongoing—but the direction is clear: when the limits of glass are approached, optical fibers must exist in a different way. Those who are the first to lay the first optical cable on the new road will determine the speed of information flow for the next fifty years.