The year is 2024. In a desert built outside Phoenix, a construction crew hits a literal wall. They are pulling fiber through conduits spec’d 18 months ago for a standard hyperscale facility. But the plan has changed: this is now a GPU-dense AI hub. Mid-pull, the conduits are full. Not due to a math error, but because the industry’s 250-micron "gold standard" fiber is too thick for the AI era.

This is the exact space crunch facing the architects of the AI revolution. We talk incessantly about the brilliance of large language models (LLMs) and the staggering compute power of new graphic processing units (GPUs), but we rarely talk about the physical "pipes" – the fiber optic cables – that must carry this data. As data centers evolve to meet the insatiable hunger of AI, we are hitting a literal wall.

The solution isn't just faster light; it’s thinner glass.

For decades, the 250-micron fiber coated diameter has been the industry gold standard. It was robust, easy to handle, and a perfect fit for the existing infrastructure.

AI has changed the maths of data center builds.

AI workloads are not like traditional cloud computing; they require massive, parallelized clusters where thousands of GPUs must talk to one another with near-zero latency. This requires a massive increase in fiber density. However, the ducts, trays, and conduits connecting outside data centers and inside our data centers are fixed assets. We cannot simply make them larger.

McKinsey's analysis finds that around 70% of total demand for data centre capacity by 2030 will be for AI-ready workloads, with demand for AI-ready capacity rising at an average rate of 33% per year between 2026 and 2030. These are environments that demand very high fiber counts that would have seemed unimaginable even five years ago.

Where Is the gap

The technical requirements of AI aren’t being met because we are fighting a battle of wavelengths and physical volume. As we move toward 1.6 Tb/s (1.6T), we aren’t just sending faster light; we are sending it over more fibers. Standard cables are becoming too large to bend, too heavy for trays, and too thick for enclosures.

The solution: smaller diameter fibers, and a lot more of them.

The industry is pivoting from 250 micron diameter to 200, 180, and now 160 microns diameter.

• The 200-micron advantage: It allows us to fit a 6,912-fiber cable into a standard conduit that previously could only hold half that capacity.
• Future-proofing: You don’t have to rip and replace installed buried cable infrastructure – merely replace the older cables with the new vastly higher capacities.

Beyond the diameter: The rise of multicore and hollow-core fiber

While shrinking the fiber solves the space problem, the sheer volume of AI data is forcing us to rethink the very architecture of the glass strand itself. This is where the roadmap for AI infrastructure moves from miniaturization to true innovation.

• Multicore fiber (MCF): If 160-micron fiber is about packing more "wires" into a pipe, MCF is about putting more "lanes" into a single wire. Historically, a fiber optic strand has a single core through which light travels. MCF embeds multiple cores – effectively multiple independent data channels – within a single strand of glass. MCF allows us to multiply data throughput by four, seven, or even more, without increasing the thickness of the cable. For the massive 1.6T and 3.2T speeds required for future AI clusters, MCF will likely become the standard "backbone" of the data center.
• Hollow-core fiber (HCF): When thousands of GPUs are waiting for a data packet to arrive before they can proceed to the next calculation, every microsecond counts. Standard fiber optic cables send light through solid glass. HCF replaces the solid glass center with an air-filled structure. Because light travels faster in air than in glass, HCF can shave critical microseconds off the time it takes for data to move between nodes. In a massive AI cluster, this "speed of light" advantage effectively "shortens" the physical distance between servers, making the entire system operate as a single, more efficient brain.

The transition to 160-micron glass, MCF, and HCF technologies is an economic imperative. In metropolitan centers like New York, London, or Tokyo, the cost of digging new trenches is astronomical, often exceeding hundreds of dollars per foot.

By treating fiber diameter and architecture as strategic variables, we move to a world of "intelligent" infrastructure.

AI demand does not wait for infrastructure to catch up. The conduits poured in concrete today must support traffic volumes that won't peak for a decade. Fiber diameter is no longer a commodity specification; it is a strategic variable.