Table of Contents
Noah Taylor, head of market intelligence and growth strategy at AFL, said the physical-layer challenge is increasingly about how much fiber can be deployed within existing infrastructure
In sum – what to know:
Fiber counts surge – AFL’s Noah Taylor says fiber counts are often increasing 5–10× or more, with some large AI campuses approaching an order of magnitude higher.
Density becomes critical – Higher fiber counts are putting pressure on conduit space, handholes, closures, and field splicing, increasing the importance of high-density cable designs.
Multi-core gains ground – As single-mode fiber pairs approach their practical limits even with 800G and 1.6T optics, Taylor says introducing multiple cores is the “next logical evolution.”
The rapid expansion of AI data centers is changing fiber requirements not only inside facilities but across the outside plant, with dramatically higher fiber counts putting new pressure on cable density, conduit space and installation methods.
Noah Taylor, head of market intelligence and growth strategy at AFL, said the physical-layer challenge is increasingly about how much fiber can be deployed within existing infrastructure. “Density is the biggest physical-layer bottleneck,” Taylor said.
The change is particularly visible in the fiber requirements of AI data centers. High-density GPU clusters require substantially more connectivity than traditional cloud environments, pushing data center-style fiber densities into metro and long-haul networks.
“The high-density GPU clusters inside data centers are pushing their cabling requirements outward into the OSP environment,” Taylor said. “What used to be data-center-only thinking—extreme density—is now required in the long-haul and metro fiber plant.”
Taylor estimates that fiber requirements are increasing by multiples rather than incremental amounts. “The increase is dramatic—often 5–10× or more, and in some large AI campuses it’s approaching an order of magnitude higher,” the AFL executive said.
That increase creates practical problems for operators building and connecting AI campuses. Traditional cable designs, conduit capacity, handholes, and closures can all become limiting factors as fiber counts rise.
Taylor said high-density cable designs can help operators accommodate more fibers without proportionally increasing the physical footprint of the outside plant. He cited AFL’s Wrapping Tube Cable as an example of a high-density approach, saying it can accommodate up to 13,824 fibers in a single cable while reducing cable diameter and the size requirements for handholes and closures.
The challenge is not simply fitting more fibers into a cable. Installation and maintenance processes also have to accommodate the higher densities. “The biggest practical challenges are splicing time, conduit space, and the physical limitations of traditional cable designs,” Taylor said.
Higher fiber counts can increase the amount of work required in the field, making installation methods increasingly important as AI infrastructure deployment accelerates. Taylor said modular connectivity and pre-terminated systems can help address some of those challenges by moving more work away from the field.
“Modular connectivity and pre-terminated cable speed up buildout. High-density cables are difficult to splice in the field, but factory connectorization simplifies installation for onsite teams at data centers,” Taylor said.
The density challenge is also extending inside AI facilities. Taylor said traditional cloud environments generally required much lower fiber counts, while AI GPU clusters can require hundreds or thousands of fibers per rack.
That is driving demand for smaller cables, higher packing efficiency, and higher-density connector technologies, according to Taylor.
At the same time, the industry is looking beyond simply adding more fibers to conventional single-mode architectures. As individual fiber pairs approach their practical limits, Taylor said multi-fiber and multi-core approaches are becoming increasingly important.
“Multi-fiber architectures are quickly becoming indispensable, especially as single-mode fiber pairs reach their practical limits—even with 800G and 1.6T transceivers,” he said.
The progression toward higher density has already involved reducing the physical dimensions of fiber and cable components. Taylor pointed to the evolution from 250-micron cladding toward 200-micron and 180-micron designs as examples of efforts to increase packing efficiency.
“Introducing multiple cores is the next logical evolution,” he said. Moving toward multi-core architectures, however, also creates new requirements for the supporting infrastructure. Splicing, testing, closures, polarity, and technician training all have to adapt to the technology.
For Taylor, the underlying direction is clear: AI is pushing fiber infrastructure toward much higher densities, both inside data centers and across the networks that connect them. The challenge is increasingly about how much connectivity can be physically deployed, managed, and maintained within the available infrastructure.
The interview with AFL’s Noah Taylor is part of a report published by RCR Wireless News and RCRTech, titled Scaling Optical Networks for the Hyperscale and AI Era, which can be accessed by clicking here.
