As AI rapidly transforms the digital landscape, the United States is expected to experience at least a threefold increase in hyperscale data center capacity by 2029. However, this capacity increase would require doubling the length of fiber routes and multiplying the total fiber mileage by 2.3. The FBA report describes how fiber optic interconnections have evolved from a “hidden” element enabling data centers to operate to a critical infrastructure for AI performance, scalability, and security.
“The need for fast, secure, low-latency connections between hyperscale data centers has accelerated with the rise of AI. Given the laws of physics, fiber remains the superior interconnect cable medium used between data centers and users,” says Deborah Kish, vice president of Workforce Research and Development at the Fiber Broadband Association. “This research reveals the magnitude of the challenge and the urgent need for new investments, policy modernization, and smarter infrastructure planning to keep pace with AI innovation.”
The FBA document identifies three key areas where interconnections are needed to unlock the full potential of AI:
New short-haul interconnections for each new hyperscale data center, averaging 135 miles of connectivity per site;
upgrades to existing long-haul fiber interconnections to increase capacity;
and entirely new long-haul interconnection routes to meet latency and bandwidth requirements.