Optical fiber form innovation: Towards lower latency and higher density

Today's optical fiber medium is already a remarkable feat in engineering. Just imagine, a single-strand optical fiber laid at the bottom of the Pacific Ocean over 20 years ago can now carry a traffic volume of 1.2Tbps, while shorter lines can even carry as much as 1.6Tbps. The fiber-to-the-home service built around 100Mbps rate at the beginning of this century is now being upgraded to 25G and 50G passive optical networks, and will support 200G PON in the next upgrade cycle. Optical fibers are continuously providing increasing speeds at lower costs, with lower latency, and in a highly reliable, robust and secure manner.

 

CableLabs believes that the optical fibers currently "deployed in the ground" could support rates of up to 50,000Gbps at some point in the future, but nowadays a large number of users already hope to raise the utility, density and performance of optical fibers to new heights.

 

Smaller-sized optical fiber options

 

At present, there are multiple paths to enhance optical fiber technology that are quietly advancing. One of the paths is to make the physical size of optical fibers smaller. The traditional single-mode optical fiber has a diameter of 242 micrometers, which is already very small. In contrast, the diameter of a human hair is approximately 50 to 100 micrometers.

 

Nowadays, companies like Corning are already capable of providing single-strand optical fibers with a diameter of 200 micrometers. This minor change can quickly have a significant impact. From residential to enterprise applications, a smaller size is always better because it enables installers to lay more optical fibers in more places, increase the number of optical fibers in capacity-tight pipelines, reduce the burden of air deployment, and make it easier for already cautious optical fiber deployment to enter offices and multi-household units.

 

The area where smaller diameter optical fibers truly shine is in AI data centers. The high-density computing required to build the next generation of AI makes every cubic inch of space precious, as racks, servers, and an increasing number of individual chips all need their own dedicated communication channels.

 

The only drawback of small-diameter optical fibers is that they need to be spliced with existing large-diameter optical fibers. This requires some specialized tools, and fiber optic technicians need to receive relevant operation training, but this is not a major challenge.

 

The rise of hollow optical fibers and multi-core optical fibers

 

Hollow core fiber (HCF) represents the next major advancement in fiber media, guiding lasers through air or vacuum rather than glass. In short, the transmission speed of light in glass is slower than that in hollow pipes (waveguides), and glass also limits the number of optical frequencies available for data transmission. If single-mode optical fibers are compared to standard highways, then hollow-core optical fibers are like highways. They can reduce latency, increase the current transmission distance, and have the potential to further enhance speed in the future.

By using hollow optical fibers, the transmission speed can be increased by 47% and the latency can be reduced by 33%. Furthermore, lower signal loss means that fewer Repeaters are required within any given distance, which translates into lower energy consumption. Over the past five years, the initial quantity of HCF has been produced and deployed to reduce latency between short-distance offices or data centers. Meanwhile, manufacturers have been constantly improving this medium to enhance its loss characteristics, making it reach or exceed the level of traditional optical fibers.

 

In 2022, Microsoft acquired Lumenisity, a manufacturer of hollow-core fibers. The company then began to produce hollow-core fibers in the UK and advanced further research on HCF. Last year, the company announced that it would deploy 15,000 kilometers of hollow fiber in its Azure data center network within two years to support the connection requirements of artificial intelligence. This year, Microsoft announced that it has successfully developed hollow-core optical fibers with better loss characteristics than traditional optical fibers, which actually opens the door to large-scale production.

 

But this did not stop. In late September 2025, Microsoft announced that it was collaborating with Corning and Heraeus Covantics to establish additional "industrial-scale production of hollow fiber" to meet the demand for this material in its data centers. We can expect that other optical fiber manufacturers will start to increase their production of hollow-core optical fibers and promote the use of this medium in data centers and other applications.

 

It is certain that there is still much work to be done to bring hollow-core optical fibers to the mainstream. This requires cultivating a group of optical fiber technicians who are proficient in the operation and splicing skills of this medium, formulating new tools and standards, and weighing the pros and cons of using standard optical fibers against the better-performing but more expensive hollow-core optical fibers. Microsoft is collaborating with Corning and Heraeus on all these issues as part of its efforts to build a standardized global ecosystem to support the large-scale deployment of hollow-core fibers in operator environments.

 

Hollow optical fibers are also of great significance to quantum computing. They can extend the transmission distance of qubits without the need for additional devices such as routers or Repeaters, which currently do not exist in quantum networks. When such devices are created, their initial cost will be higher than that of existing network devices. Hollow-core optical fibers should be able to reduce the demand for future quantum network equipment, thereby saving money and accelerating deployment time.

 

Multi-core fiber (MCF) is the third path for optical fibers to create greater bandwidth. It places multiple optical fiber cores into a single optical fiber, enabling more signals to be transmitted simultaneously along a single fiber. Multi-core optical fibers increase fiber density and bandwidth. Manufacturers such as Lightera and Sumitomo Electric are working on improving them and commercializing them for widespread use.

There have already been some very remarkable demonstrations of multi-core technologies. Earlier this year, Sumitomo and the National Institute of Information and Communications Technology of Japan announced a world record. They used 19-core optical fibers to transmit over 1PB per second of data over a distance of more than 1,800 kilometers (equivalent to the distance from Missouri to Montana). Closer to deployment is that Lightera is sending samples of multi-core optical fiber solutions to selected customers and has demonstrated its ability to produce optical fibers ranging from 4 to 8 cores.

 

Multi-core optical fibers have a wide range of applications, including submarine and land connections, as well as high-density, high-speed connections between switches, servers and storage devices in data center applications. Lightera has demonstrated its ability to support 8-core multi-core fiber at 800Gbps in short-range applications and 4-core multi-core fiber at 400Gbps over a distance of 10 kilometers.

 

However, like hollow-core optical fibers, multi-core optical fibers also face their own challenges. Although there is a multi-core fiber working group within the Advanced Photonics Coalition, they have not yet established standards regarding basic characteristics such as the number of cores, core layout, and cladding diameter, which makes the current field deployment of each multi-core fiber a customized project. Special tools for multi-core splicing need to be built, especially to ensure that they can be spliced quickly, with low loss and high strength. Finally, well-trained multi-core fiber optic technicians are also needed, and it is best for them to follow and abide by the established standards.

 

Despite this, as multi-core fibers mature, they will keep pace with standard fibers and hollow-core fibers, providing more network options for data centers, hyperscale providers, cloud and service providers, and enterprises. Perhaps the most forward-looking observation I can offer is that network planners should carefully consider the future balance between the deployed optical fibers and the available pipelines, so as to be able to introduce new solutions when customers need them (such as hollow or multi-core optical fibers).

Traditional single-strand optical fibers will not disappear, but it is always a good thing that they can offer options for advanced users who pursue lower latency, higher density and/or greater bandwidth.