Zayo plans 5,000 fiber infrastructure route-miles to meet AI demand

The continued growth in artificial intelligence workloads is creating unprecedented demand for fiber infrastructure, with orders for fiber connections increasing by orders of magnitude in the past 18 months. Zayo, a major telecom infrastructure provider, has announced plans to build over 5,000 fiber route-miles specifically to support AI-related data traffic needs.

Current market dynamics: The scale of fiber infrastructure demands for AI applications has grown dramatically, shifting from typical orders of 8-12 fibers to requests for 144-432 fibers per installation.

• Zayo has secured more than $1 billion in AI-related deals in 2024, with an additional $3 billion in their pipeline
• The company remains the only provider building long-haul routes at scale in the United States
• Current fiber infrastructure largely relies on installations from around 2000, creating a potential capacity gap

Strategic expansion plans: Zayo’s five-year development strategy includes five new routes and seven overbuild projects to create direct, low-latency connections between key data center hubs.

• New routes will connect major tech corridors including Columbus to Indianapolis and Atlanta to Ashburn
• Overbuilds will enhance capacity on critical paths like Denver to Dallas and Salt Lake City to Reno
• Route selection was based on detailed analysis of chip manufacturing trends, power availability, and customer demand forecasts

AI workload considerations: Different types of AI applications will drive distinct patterns in fiber infrastructure requirements.

• Training workloads will likely concentrate in areas with abundant power resources, such as West Texas
• Inference operations are expected to generate more bandwidth demand through data center-to-data center and data center-to-consumer communications
• Back-end processes for AI applications will require extensive metro communications and interconnection

Infrastructure placement factors: Data center location decisions have evolved to prioritize power availability above other considerations.

• Traditional factors included population density, connectivity options, and power costs
• Power has become the primary constraint due to its scarcity and the high cost of transmission infrastructure
• Building fiber routes to power sources is 4-10 times less expensive than constructing new power transmission lines

Market competition outlook: The surge in AI-driven demand may attract new players to the fiber infrastructure market, though significant barriers to entry exist.

• The long-haul fiber business requires specialized expertise and substantial capital investment
• Historical attempts by smaller providers often ended in acquisition, as evidenced by Zayo’s growth through dozens of company purchases
• The industry must relearn large-scale fiber deployment skills after a two-decade lull in major construction

Looking ahead: The power equation: The intersection of power availability and fiber infrastructure will likely shape the future of AI deployment, with organizations optimizing for power access rather than traditional connectivity considerations. This shift could lead to new patterns in data center placement and potentially create opportunities for regions with abundant power resources to become major AI compute hubs.