By Shane Snider
Senior News Writer, Data Center Knowledge
August 21, 2026
Executive Overview
As the artificial intelligence revolution continues to strain the physical limits of global technology infrastructure, telecommunications provider Zayo is taking proactive measures to prevent severe supply bottlenecks. The company has officially reserved a material portion of its anticipated optical fiber manufacturing capacity with Corning through 2030.
This multi-year supply agreement is designed to safeguard Zayo’s ambitious network expansion plans, which include the deployment of 15,000 route miles of new and overbuilt infrastructure across North America. Driven by the staggering data transfer requirements of machine-to-machine AI workloads, the partnership underscores a shifting industry reality: high-count fiber optic cabling has rapidly emerged as a critical physical constraint, standing right alongside power availability and real estate as core limiting factors for modern data center campuses.
Detailed Chronology: The Strategic Road to 2030
Anticipating the Upcoming Supply Crunch
Historically, major telecommunications operators like Zayo procured fiber strictly on an as-needed project basis. Although long-haul network construction typically requires years of meticulous planning, raw materials were traditionally ordered closer to the actual deployment phase. However, the unprecedented velocity of the generative AI boom has fundamentally altered supply chain dynamics.
While Zayo has not yet experienced acute, material fiber shortages across its active worksites, company leadership recognized early warning signs of systemic strain. Troy Lupe, Chief Network Officer at Zayo, noted that the staggering volume of projected network demand leaves little room for complacency. Because building advanced optical manufacturing lines and securing upstream raw materials takes years, Zayo elected to secure its supply chain visibility years in advance.
By locking in manufacturing capacity with Corning through 2030, Zayo has transformed its procurement strategy. Instead of reacting to market shortages, the telco is working hand-in-hand with Corning to ensure that manufacturing lines are actively opened and raw chemical supplies are secured ahead of schedule.
Unpacking the 15,000-Route-Mile Expansion
Zayo’s broader funded network expansion encompasses 28 distinct routes—either newly proposed or currently in progress—consisting of 10 completely new long-haul routes and 18 overbuild projects. According to official network statistics released following discussions with Lupe, these combined initiatives account for:
- 15,000 new or overbuilt route miles.
- 20 million new or overbuilt fiber miles.
A significant catalyst for this massive buildout is Zayo’s high-profile strategic partnership with Nvidia. Out of the 28 total routes slated for construction, 24 are tied directly to the Nvidia infrastructure initiative. This specific program represents 8,000 route miles and 15 million fiber miles.
Zayo has already successfully delivered four routes this year, with an additional four expected to reach completion before the close of 2026. While some early capacity is already coming online, the bulk of the Nvidia-backed infrastructure will be systematically rolled out over the next several years.
Supporting Context & Metrics: The Optics and AI Nexus
The Anatomy of AI-Driven Network Demands
Industry analysts emphasize that while AI does not necessarily require a fundamentally different type of fiber optic glass, the scale, density, and topology of the network have changed radically.
Jimmy Yu, Vice President at Dell’Oro Group, points out that the primary differentiator for AI infrastructure is the sheer volume of fiber packed into every conduit. Whereas traditional enterprise and consumer traffic relied on lower fiber counts, AI data centers require hundreds—if not thousands—of fibers per conduit to facilitate multi-terabit interconnectivity between distributed facilities.
Furthermore, AI workloads are characterized by intense machine-to-machine communications. As Troy Lupe observed:
"It’s machines talking to machines, and they’re transferring a lot more data than humans talking to humans or humans talking to machines."
To manage this massive data throughput without latency degradation, network operators are forced to deploy advanced architectural strategies:
- Route Diversity ("Triversity" and "Quadversity"): AI operators require multiple physically diverse paths between network endpoints to guarantee zero downtime during unexpected fiber cuts.
- Direct Corridors: Minimizing add/drop points along long-haul routes reduces latency between distant data center clusters.
- Geographic Expansion: Because modern AI data centers require unprecedented blocks of power, campuses are frequently built outside traditional urban telecom hubs, necessitating extensive new north-south and east-west long-haul routes.
Market Momentum and Corning’s Expansion
The broader optical transport market reflects these hyper-growth trends. According to Dell’Oro Group data:
- Optical transport equipment revenue grew 15% year-over-year in the second quarter.
- Data center interconnect (DCI) revenue—driven by WDM systems and IPoDWDM optics—surged by an impressive 45% year-over-year.
Corning, meanwhile, has positioned itself as one of the primary beneficiaries of the physical AI infrastructure buildout. In addition to securing massive optical supply agreements with hyperscalers like Meta and Amazon, Corning partnered with Nvidia to aggressively expand its U.S. optical manufacturing capacity. Driven by these wins, Corning’s Optical Communications business experienced a staggering 36% year-over-year growth in the first quarter alone.
Although a Corning spokesperson declined to disclose specific financial or volume details regarding the Zayo agreement, the manufacturer emphasized its long-standing policy of keeping production facilities geographically close to customers—exemplified by its massive optical fiber and cable manufacturing plants in North Carolina.
Official Statements and Industry Insights
The convergence of AI workloads and physical network constraints has prompted industry experts to reevaluate how network operators plan for the future.
Ron Westfall, Vice President and Practice Lead for Networking and Infrastructure at HyperFrame Research:
"High-count fiber optic cabling has rapidly emerged as a secondary, structural constraint alongside power. The multi-terabit connectivity required between distributed AI facilities means that geographically diverse routes are no longer a luxury—they are a fundamental operational requirement."
Jimmy Yu, Vice President at Dell’Oro Group:
"I think Zayo is being proactive in building the fiber network that will be needed for future AI infrastructure in the US. The agreement with Nvidia for 8,000 of those 15,000 route miles gives the company immense confidence in future capacity demand."
Troy Lupe, Chief Network Officer at Zayo:
"We have not really experienced material issues yet, but if you look at the amount of demand coming, it takes time to build factories across the stack. We are reserving capacity and planning with Corning so that they are opening manufacturing lines and securing supplies to deliver all of this fiber."
Future Outlook: Navigating Permitting Risks and the Inference Shift
Execution Risks: The Permitting Hurdle
Despite having secured a stable, multi-year supply of optical fiber through Corning, Zayo still faces complex logistical hurdles. Lupe noted that permitting and construction across multiple municipal, state, and federal jurisdictions represent the single biggest execution risks moving forward.
While Zayo has established a robust track record of deploying infrastructure across many of these regions, local right-of-way laws and environmental reviews vary wildly. Navigating these regulatory frameworks will dictate the exact speed at which the remaining 15,000 route miles materialize.
The Looming Impact of AI Inference
Looking ahead beyond 2030, network architectures may undergo yet another transformation as AI workloads pivot away from training and toward large-scale inference.
Currently, the vast majority of capital expenditure and data center construction is dedicated to training foundational models. However, as generative AI applications become ubiquitous, consumer-facing inference and personalized real-time content generation will scale exponentially.
While much digital content today can be efficiently cached at the network edge close to end-users, heavy AI-generated compute queries could permanently alter traffic patterns inside metropolitan areas. Should inference decentralize further, the demand for hyper-dense metro connections and resilient long-haul transport will only accelerate.
By proactively securing manufacturing capacity with Corning today, Zayo has insulated its capital deployment strategy against macroeconomic volatility, ensuring that America’s burgeoning AI infrastructure has the robust digital highways it needs to thrive well into the next decade.
