The agreement secures a material portion of the fiber Zayo needs for its 15,000 route-mile network expansion through 2030.
Telecommunications firm Zayo is reserving manufacturing capacity with Corning for a material portion of the fiber it expects to need through 2030, protecting plans to build 15,000 route miles of new and overbuilt infrastructure as AI drives demand for data center connectivity.
The agreement covers part of the telco’s broader funded network expansion, which includes 28 new or in progress routes, 10 new routes, and 18 overbuilds. The projects represent 15,000 new or overbuilt route miles and 20 million new or overbuilt fiber miles, according to network statistics Zayo provided after a Data Center Knowledge interview with Chief Network Officer Troy Lupe.
Zayo has not experienced a material fiber shortage, but Lupe said the company is seeing signs of future supply pressure as network demand grows.
“I think we’ve seen some concerning signals,” Lupe said. “We have not really experienced material issues, but I think that, if you look at the amount of demand coming, the amount of historical demand needed, it takes time to build factories across the stack.”
Historically, Zayo ordered fiber as projects required it, although long-haul construction itself takes years of planning. The expanded Corning agreement provides Zayo with supply visibility through at least 2030 and enables the two companies to plan manufacturing capacity around Zayo’s projected requirements.
“We’re reserving capacity with them,” Lupe said. “We’re planning with them so that they’re opening the manufacturing lines and buying the supplies they need to deliver all of this fiber.”
The agreement comes as Corning has secured a growing role in the physical infrastructure behind AI networks. The company has signed major optical supply agreements with Meta and Amazon and partnered with Nvidia to expand US optical manufacturing capacity. Its Optical Communications business grew 36% year over year in the first quarter.
A spokesperson for Corning said the company is not disclosing additional details about the customer agreement but maintains a policy of manufacturing close to customers to support reliable supply and service.
Corning operates two of the world’s largest optical fiber and cable manufacturing facilities in North Carolina, reflecting its significant manufacturing presence in the United States.
Addressing Fiber Constraint #
Ron Westfall, vice president and practice lead for networking and infrastructure at HyperFrame Research, said AI infrastructure is making fiber availability a physical constraint alongside power and other infrastructure needed to build data centers.
“High count fiber optic cabling has rapidly emerged as a secondary, structural constraint,” Westfall told Data Center Knowledge, pointing to the multi terabit connectivity required between distributed AI facilities and data centers.
Westfall also pointed to geographically diverse routes as a growing requirement as AI campuses move into markets outside traditional telecommunications hubs. Long term fiber commitments can protect network construction schedules as operators plan projects years ahead.
Jimmy Yu, vice president at Dell’Oro Group, said Zayo is building ahead of demand it can now see through its Nvidia agreement, which covers 8,000 of the company’s 15,000 planned route miles.
“I think Zayo is being proactive in building the fiber network that would be needed for the future AI infrastructure needs in the US,” Yu said. “I’m sure the agreement with Nvidia for 8,000 of that 15,000 route miles gives the company added confidence in the future demand for capacity.”
Yu said AI does not require a fundamentally different long-haul network, but the scale and design of the infrastructure are changing.
“The main difference is that the amount of fiber per conduit will need to be in the hundreds, if not thousands,” Yu said.
AI networks also require more diverse paths for resiliency, direct routes to reduce latency between data centers and fewer add/drop points along routes, Yu said.
The broader optical market is already reflecting the demand. Optical transport equipment revenue grew 15% year over year in the second quarter, while data center interconnect revenue from WDM systems and IPoDWDM optics grew 45%, according to Dell’Oro. Yu said demand is strong between neighboring data centers and across long distance backbones, particularly in North America.
Building 15,000 Route Miles #
Zayo’s broader expansion spans engineering, permitting and construction. The company’s work with Nvidia accounts for 24 of the 28 routes in the buildout. That program includes six new routes, 10 overbuilds and eight existing fiber routes, representing 8,000 new or overbuilt route miles and 15 million new or overbuilt fiber miles.
Lupe said Zayo has delivered four routes this year and expects to deliver another four during the remainder of 2026. The Nvidia projects are at various stages, with much of the work in engineering and several in the environmental and permitting phases.
Some capacity is being delivered this year, with most of the Nvidia build expected over the following several years. Permitting and construction across multiple jurisdictions pose the biggest remaining execution risks, Lupe said.
Zayo has already built or overbuilt infrastructure in many of the jurisdictions involved, but requirements vary by right of way and permitting authority.
With fiber supply secured, Zayo can focus on those remaining construction requirements.
AI Expands Route Diversity #
AI traffic does not require a fundamentally different type of fiber. The difference is the volume of traffic generated by machines communicating with other machines.
“It’s machines talking to machines, and they’re transferring a lot more data than humans talking to humans or humans talking to machines,” Lupe said.
Zayo is seeing greater demand for physically diverse network paths between critical facilities as AI operators seek protection against outages.
Lupe described the requirement as demand for “triversity” and “quadversity,” referring to multiple diverse routes between network endpoints.
New compute facilities are also taking Zayo into markets where it historically would not have built, following the availability of power. That is creating demand for north-south and east-west long-haul routes, particularly where existing corridors have limited capacity or physical diversity.
Yu said Zayo’s announcement is part of broader network construction by communications service providers supporting the AI buildout.
“As more new data centers get built, more new fiber routes will also be deployed,” Yu said.
Inference Could Move Capacity #
The network requirements could change as AI workloads move from training toward inference.
Most current AI infrastructure investment is associated with training, Lupe said. Zayo is beginning to see inference activity inside metropolitan areas, but personalized AI-generated content could increase traffic between users and compute facilities.
Much digital content today can be cached relatively close to users. AI-generated content could require more connections between users and the locations where computing is available.
“If that potentially starts to change, that could materially change the demands inside the metros,” Lupe said.
That possibility gives Zayo another reason to plan for capacity and route diversity as it builds its long haul network through 2030.