Large scale enterprises have clearly stated why hollow fiber optics attract them: AI is forcing data centers to be distributed further, while not allowing delays to disrupt network design. But Badri Gomatam, CTO of Sterlite Technologies Ltd., believes that the industry should move beyond the narrative of delay.

Gomatam told Fierce Network that one of the most important advantages of hollow core fiber is its ability to tolerate higher optical power without the same level of interference as traditional solid glass fibers. Gomatam said, “When you transmit in air, the interaction between light and air is not as strong as between light and glass.” This reduced resistance means “you can inject much more energy,” thereby extending the transmission distance, he added.

Why is power important

Here is why increasing power levels in hollow core fibers is more effective than attempting the same approach in single-mode fibers. In standard single-mode fibers, adding more wavelengths and higher transmission power will ultimately create bottlenecks in the form of crosstalk. That is to say, channels begin to interfere with each other, and even if enough light reaches the receiving end, errors will occur. Hollow fiber optics have not eliminated physical laws, but Gomatam stated that this shape significantly reduces the limit that operators can push a single fiber optic cable.

This is important because the construction of AI data centers not only creates a demand for faster links, but also increases the pressure on every part of the optical transmission stack – bandwidth, distance, power, pipeline space, and the ability to connect more buildings without constantly rebuilding physical facilities.

AWS has stated that hollow core fiber can provide about 30% delay improvement compared to standard fiber and help increase transmission distance by nearly 50% while maintaining the same delay target. This cloud giant is already deploying the technology to connect about 10 data centers, but it told Fierce that supply is still a limiting factor. Microsoft has also taken measures to expand the scale of hollow fiber optics, reaching agreements with Corning and Heraeus Covantics to increase global HCF production for Azure. Corning stated that as part of the collaboration, its North Carolina fiber optic and cable manufacturing facility will produce Microsoft’s hollow fiber optic cables.

Manufacturing issues

However, currently, the manufacturing of hollow core optical fibers is much more difficult than standard optical fibers. Gomatam describes the basic structure as “glass tubes nested within glass tubes”. This brings precision manufacturing challenges: when the preform is heated, drawn, and cooled, the internal hollow structure must maintain its shape and size. He said, “If it’s hollow, these small nested structures must maintain their position during the drawing process. This means that the drawing speed of each tower will be slower than today’s solid fibers.”

Gomatam is not the only one pointing out this challenge. AWS’s Matt Rehder told Fierce that “the real challenge is manufacturability,” including low yield and high manufacturing costs for fiber optic cables of sufficient length to connect multiple data centers.

Hollow core and multi-core

STL is interested in both hollow core and multi-core fibers, but Gomatam frames them to solve different problems on different timelines. He said that multi-core fiber is more likely to be certified by customers in the near future because it is closer to the fusion and connection methods already known by operators. Gomatam said, “Persuade someone to deploy multi-core faster.” He was referring to space, weight, and pipeline capacity limitations. Within the data center, multi-core is a great story, “Gomatam added, adding that the foundation of multi-core is also more solid because the underlying standard work already exists, while the hollow core standard is still developing.

In contrast, there are stronger reasons for using hollow cores outside of data centers, where transmission distance and latency can change the economics of placing buildings for operators and large-scale enterprises. But it also comes with a steeper operational learning curve, including fusion, training, termination, and connector ecosystems that still need to mature.

Gomatam stated that knowledgeable clients are asking the right questions: when will the technology be ready, when will the cost structure improve, and where is the reasonable total cost of ownership. But he added that standardization may be particularly important if Airchip is to move from deployment led by ultra large scale enterprises to a broader telecommunications ecosystem.