AI infrastructure workloads are exponentially increasing fiber connectivity requirements across data centers

Infrastructure company Zayo reports that artificial intelligence training and inference operations are driving unprecedented demand for fiber connectivity, with individual customer requirements escalating from dozens of fibers to thousands. The geographic distribution of AI workloads across compute clusters and data centers is fundamentally reshaping network architecture, and demand for dark fiber in major metros has multiplied by as much as 20 times. Unlike previous capacity buildouts, this expansion cycle comes after the market has fully absorbed prior infrastructure investments, reducing risks of oversupply.

The shift in fiber demand reflects AI's unique infrastructure footprint. Training large language models and running inference operations at scale require massive, sustained bandwidth between distributed computing facilities. Unlike previous technology cycles driven by speculative investment, this expansion is grounded in concrete deployments with long-term customer commitments concentrated in specific metropolitan areas where infrastructure gaps currently exist.

Beyond fiber cables themselves, the construction challenge has become multifaceted. Permitting across hundreds of jurisdictions, securing skilled construction labor, and coordinating complex logistics now rank alongside fiber availability as limiting factors. Zayo's experience suggests that project sequencing and execution quality could become as critical as raw material availability in determining how quickly AI infrastructure can scale.