AI Workloads Have Doubled Long-Haul Fiber Demand
Distributed AI clusters are exhausting 2000s-era network capacity, creating a bottleneck for new data center connectivity.
Updated on Oct. 6, 2026 in Data Centers
Live Poll
Is now a good time to invest in major national digital infrastructure upgrades?
Zayo reports that global demand for long-haul dark fiber doubled between 2024 and 2025 as AI training and inference requirements scale. This surge has fully utilized the network capacity established during the investment wave of the 2000s.
Why it matters
The shift toward distributed AI compute clusters demands unprecedented physical connectivity between data centers. This surge in traffic is forcing providers to confront significant construction delays related to local permitting and skilled labor availability.
Metro-area dark fiber demand has surged by 20x compared to historical baselines, while long-haul demand has doubled in just one year. This consumption has fully exhausted capacity built during the network infrastructure expansion of the 2000s.
The players
Zayo
A global provider of fiber-based communications infrastructure, including dark fiber and long-haul network connectivity for data centers.
The details
Dark fiber — unused optical fiber cables already installed but not yet lit with network equipment — is being activated to connect massive compute clusters. Connecting these sites requires navigating complex municipal rights-of-way and regulatory environments, where construction delays at any stage ripple across the entire project timeline. Zayo has addressed this by delivering new long-haul routes over the past five years to meet the thousands of individual fibers now required by single customers.
Timeline
2000s: Initial network infrastructure investment wave occurred.
2021-2026: Zayo delivered new long-haul fiber routes.
2024-2025: Long-haul dark fiber demand doubled globally.
The Tech Race
The current surge in fiber demand marks the total exhaustion of the network capacity established during the 2000s investment wave. Infrastructure providers are now in a race to outpace local regulatory and labor constraints to support the massive connectivity needs of modern AI clusters.
Data center operators and enterprises should expect longer lead times for new network deployments due to municipal permitting and skilled labor bottlenecks. Organizations will increasingly need to plan physical supply chain requirements years in advance to secure connectivity.
The takeaway
The rapid scaling of AI is creating a physical bottleneck that software optimization cannot circumvent. Watch for future reports on municipal regulatory reforms or new labor-efficient construction methods as the industry attempts to resolve the fiber shortage.
Further reading
Explore the evolving infrastructure requirements for AI in our Data Centers section.
Source note: This article includes information reported by RCR Wireless News.
Live Poll
Is now a good time to invest in major national digital infrastructure upgrades?
