Zayo, Equinix develop AI network framework

Artificial intelligence (AI) has transformed the digital infrastructure landscape over the past two years, but there has been no reference for connecting training, inference and enterprise infrastructure to accompany that transformation. To address this, Zayo and Equinix have published “the industry’s first AI Infrastructure Blueprint”.

The joint infrastructure architecture framework from the communications infrastructure provider and the digital infrastructure company sets out to define how next-generation infrastructure powers AI workloads. The companies stated that neocloud and AI providers will have a framework that clarifies the roles of high-capacity networks, interconnection hubs, training and inference datacentres, along with a clear model for connecting them.

As AI-driven bandwidth demand is expected to grow up to six times by 2030, Zayo has invested in fibre capacity and new routes required for AI-scale data movement, including a commitment to build more than 5,000 new route miles of long-haul fibre.

Bill Long, chief product and strategy officer at Zayo, said: “We’re introducing a network standard and datacentre best practices that makes AI communication infrastructure scalable, extensible and ready for what comes next.”

The AI Infrastructure Blueprint fundamentally notes that AI at scale spans training locations, distributed inference and network and interconnection nodes that link them. The AI Infrastructure Blueprint provides a infrastructure architecture framework for how these pieces fit together: Equinix as the neutral interconnection hubs for connecting networks, training and inference infrastructure; and Zayo as the high-capacity fibre and network services, linking hubs, data sources and workloads.

The result is “a clear, structured path” to launch and scale AI communication infrastructure, giving neocloud and generative AI providers practical private connectivity guidance to reduce complexity and to speed up roll-out.

The blueprint aims to deliver validated blueprints at scale, with reference designs that focus on the network elements that matter at scale, reducing trial and error and shortening time to market for AI training and inference. It offers practical direction across key network layers and elements, informed by Equinix and Zayo’s cloud connectivity and IP peering experience, to help businesses design for long-term AI growth. It also presents a shared language for the AI ecosystem using common terminology that aligns customers, partners and suppliers across the AI networking stack.

“Neoclouds and AI providers face a widening gap between their ambitions and the infrastructure needed to support them, with scaling networks among the toughest challenges,” said Craig Matsumoto, contributing analyst at Futuriom Research.

“The AI Infrastructure Blueprint brings needed clarity to the industry by mapping how Zayo’s extensive fibre backbone with Equinix’s global connectivity fabric work together to give teams a repeatable model to scale and support low latency inference at the edge. In our view, this blueprint marks a pivotal step toward a common model for scaling AI across communications infrastructure.” 

Arun Dev, vice-president of digital interconnection at Equinix, added: “As enterprises race to operationalise AI, they’re realising success depends on more than GPUs. It requires distributed, high-performance connectivity that extends all the way to the edge.

“With Fabric Intelligence, those connections become smarter with real-time discovery, activation and optimisation that reduce integration risk, accelerate scale and help ensure AI deployments evolve in step with the way the industry is moving.”