Table of Contents
Zayo chief network officer Troy Lupe told RCRTech that AI factories and hyperscale data centers are increasingly being built where power and land are available, often outside the traditional data center markets
In sum – what to know
AI changes locations – AI factories and hyperscale data centers are increasingly being built where power and land are available.
New corridors emerge – These locations can sit outside traditional data-center markets, creating demand for new high-capacity long-haul routes.
Connectivity follows compute – Zayo says distributed AI infrastructure will require connections to data centers, clouds and major network hubs.
The expansion of AI infrastructure is changing where long-haul networks need to be built, as AI factories and hyperscale data centers are increasingly being developed in locations where power and land are available, according to Zayo chief network officer Troy Lupe.
“AI factories and hyperscale data centers are increasingly being built where power and land are available, often outside the traditional data-center markets,” Lupe told RCR Wireless News.
That shift is creating demand for new high-capacity routes connecting those locations with the wider digital infrastructure ecosystem. “It is changing where long-haul fiber needs to go and creating demand for new high-capacity routes connecting those locations to other data centers, clouds, and major network hubs,” Lupe said.
The change is linked to the increasingly distributed nature of AI infrastructure. Rather than being concentrated exclusively within established data-center markets, new AI facilities can emerge in locations where the physical conditions required to support them are available.
For network providers, that creates a requirement to connect those emerging locations with the existing infrastructure around them. Zayo says its network position gives it the ability to address that requirement through a combination of long-haul and dense metro assets.
“Zayo is well positioned for that shift because we operate as a Tier 1 ISP with both long-haul and dense metro assets,” Lupe said. “That gives us the ability to connect emerging AI markets into the broader network ecosystem, from regional data-center clusters to major cloud and interconnection hubs.”
The requirement for connectivity also extends beyond the largest hyperscalers. As AI infrastructure becomes more distributed, the locations where compute is deployed increasingly need to remain connected to other parts of the ecosystem. “As AI infrastructure becomes more geographically distributed, high-capacity connectivity becomes a fundamental part of enabling it,” Lupe said.
The change also reflects the different connectivity requirements created by AI workloads. Training requires substantial bandwidth between compute clusters, data centers, and supporting infrastructure, while inference at scale is making connectivity more distributed across clouds, data centers, enterprises and end users.
That means the network requirements associated with AI infrastructure are not limited to a single connection between a data center and a major network hub. Connectivity needs to extend across the different locations and environments involved in delivering AI workloads.
For Zayo, that creates a need to build routes in locations where AI infrastructure is emerging rather than limiting network expansion to established data-center markets.
The company’s long-haul and metro footprint provides connections between emerging AI markets and the broader network ecosystem, according to Lupe. “The compute may go where the power is, but it still has to connect to the rest of the digital ecosystem,” he said.
