For many executives, fiber has been important but largely invisible. AI is changing that by turning connectivity into part of the production system itself. The strategic question is no longer how much bandwidth a business will need, but whether it can secure the right routes, with enough resilience and control, before those routes become scarce.
It is tempting to call this a capacity story. It is not. AI training and inference are changing where traffic travels, how often data is replicated, and how much performance depends on the network between facilities. Workloads now span campuses and regions, with active-active architectures reading and writing data in multiple locations at once. As a result, route diversity has moved to the top of the agenda: hyperscale and neocloud operators are seeking three, four and even five diverse fiber paths between critical sites, driving significant demand.
That shift is creating a new class of long-haul demand. The most valuable corridors are not necessarily the longest or those with the most advertised fiber, but those connecting the right data center markets with direct, low-latency paths, with the fewest interconnection (aka “failure”) points and room to expand. The quality of the catchment area now matters as much as the physical asset.
Supply is another reason this market deserves attention now. Much of the U.S. backbone was built 20 to 30 years ago, when network engineering treated population centers as the source of traffic demand and drew routes to connect them. AI demand is no longer fully aligned with those legacy routes. Many of those systems are highly interconnected and also have limited duct capacity and too few strands. Lighting DWDM capacity helps but does not remove physical constraints, latency needs and fault tolerance. In a growing number of corridors demand for dark fiber exceeds supply.
The buyers shaping this market are also raising the bar. Hyperscale and neocloud operators want speed, reliability, scalability and control: direct data center connectivity, diverse routes, clear repairability and capacity for 10 to 20 years of growth. They do not care about the service layer. They want the physical layer (dark fiber or conduit) and will build the higher layers themselves to meet their own needs. And they work to AI deployment schedules; a route that arrives late may have little value, however strong its specification.
This puts execution on the same footing as engineering. The right questions are precise. How many independent paths reach each critical facility? Can the route be delivered in the required window? Will the provider invest ahead of demand, and does the contract preserve room to expand? Dark fiber, dedicated conduit and IRUs provide control only when the commercial and operational model around them is sound. The winners will offer more than strands: a dependable infrastructure platform that works when the customer needs it.
The financing question is equally important. Long-term capacity agreements can anchor a new build, but an anchor customer does not remove every risk, and operators may need to invest ahead of contracted demand. Diligence should test five things together: the demand profile of the connected markets, the physical diversity of the route, the construction plan, the commercial terms and the expansion potential beyond the first anchor. A strong customer does not compensate for a fragile route, and abundant fiber creates no value if the corridor cannot be delivered in time.
For C-suite leaders, the decision is therefore broader than whether to build or buy. It is about which corridors matter, where ownership creates advantage, where partnerships are more efficient, and how connectivity fits the wider data center and cloud strategy. The same discipline applies to carriers, data center platforms and investors, who must separate routes with durable demand from assets that simply add exposure to a constrained market.
The fiber market is moving faster than the old planning cycles were designed to handle. Companies that wait for demand, routes and pricing to become obvious may discover that the best options have already been committed.
Richard Knowlton, Senior Advisor, PMP Strategy
About Richard Knowlton and PMP Strategy
Richard Knowlton is a Senior Advisor at PMP Strategy and brings extensive experience across telecom, digital infrastructure, investment banking and commercial diligence. In his work with investors, operators and management teams, he translates market signals into practical decisions about connectivity, data centers, fiber assets and growth strategy. PMP Strategy combines that senior perspective with deep analytical capabilities, a global platform and a network of industry experts to help clients navigate complex business challenges, evaluate opportunities and act with confidence.
