Network Expansion: Edge Computing's Infrastructure Race

For data centre operators, network expansion has shifted from adding raw bandwidth to building distributed, low-latency fabrics that can host AI inference, IoT and data-sovereign workloads as close to the user as possible. 

Analyst forecasts from IDC suggest global edge computing spend will climb from about US$261bn in 2025 to nearly US$380bn by 2028, underlining how capex is tilting towards edge-ready infrastructure. At the same time, market value for edge platforms is expected to rise from roughly US$710bn in 2026 to more than US$6tn by 2035, driven by demand for real-time analytics and AI at the periphery, according to Precedence Research.

This growth is forcing operators to rethink network design from the optical layer up. 

Rather than hub-and-spoke architectures feeding central hyperscale campuses, the emphasis is on meshed metro rings, dense backhaul and software-defined connectivity that can steer workloads dynamically between edge sites, regional hubs and public cloud regions. 

And with users now expecting sub-10ms round-trips for interactive and AI-powered services, this means pushing compute and connectivity deeper into cities and industrial zones.

AI-driven capacity and fibre densification

AI is the dominant driver behind current network expansion programmes, both for carriers and neutral colos. 

Lumen, for example, is part-way through a multibillion-dollar build to add 34 million intercity fibre miles by 2028, having already deployed more than 2.2 million new intercity fibre miles and 5.9+ Pbps of capacity in 2025 alone. 

The company reports that its 400G-enabled network now spans over 100,000 route miles, designed to deliver less than 5ms latency at the edge for up to 97% of US business demand.

Within the data centre footprint itself, operators are also upgrading east–west and cloud-on-ramps to 400G and beyond.

Lumen's recent refresh of its US data centre and cloud connectivity fabric, enabling up to 400Gbps Ethernet and IP across more than 70 data centres in 16 metros, reflects a wider trend towards treating interconnect as a first-class AI resource, not a utility. 

In parallel, players like Cisco are rolling out integrated edge platforms that combine networking, compute, storage and security in a single appliance.

Operational and commercial considerations

Network expansion at the edge is not solely a capex race. Operators also face rising energy costs and an increasingly demanding security landscape. 
As more AI inference and sensitive data moves to edge locations, embedded security in routers, virtual network functions and fabric controllers becomes non-negotiable.

Commercial models are evolving in parallel. Some carriers are using indefeasible rights of use on conduits and fibre to lock in strategic corridors, while others are leaning into consumption-based edge connectivity, allowing enterprises to spin up virtual routing, firewalls and interconnect in minutes rather than months. 

For data centre leaders, the strategic challenge is to align these models with customer expectations around predictability, while still monetising premium low-latency and AI-optimised paths that will increasingly differentiate one edge-ready facility from another.

Equinix – programmable interconnection at the edge

Equinix has emerged as a reference model for edge‑centric network expansion by combining global reach with software‑defined interconnection. 

The company now operates more than 270 data centres across 77 metros and is actively investing in new sites and expansions, with some 20 projects scheduled to open in 2026 alone. 

Its Equinix Fabric service allows customers to create private, on‑demand connections between their infrastructure, cloud providers and network services, effectively bypassing the public internet while retaining flexibility. 

Recent collaboration with Resolute CS extends that capability to the enterprise last mile by automating design, pricing and ordering of global access through the Resolute NEXUS platform, giving customers transparent access to thousands of network providers.

In 2026, Equinix is also introducing Fabric Intelligence for select customers, applying agentic automation to monitor performance and automatically adjust paths and capacity. 

This positions the company as more than a landlord: it becomes an active participant in customers’ hybrid and multicloud networking strategies. 

For data centre leaders, the lesson is clear – network expansion at the edge is increasingly about abstraction and automation, not just adding ports and fibres.

Lumen Technologies – building an AI backbone

Lumen is executing one of the most aggressive fibre‑based network expansion strategies in the market, explicitly framed around supporting AI and edge workloads. 

By August 2025 the company had deployed over 2.2 million new intercity fibre miles, on track to reach 16.6 million by the year’s end, as part of a broader goal to add 34 million intercity fibre miles by 2028. 

Over the same period, Lumen added more than 5.9 Pbps of capacity, using next‑generation fibre optic cables that can fit twice as many fibres into existing conduits.

This long‑haul expansion underpins a 400G‑enabled network that now spans more than 100,000 route miles, delivering sub‑5ms edge latency across as much as 97% of US business demand. 

Within the data centre landscape, Lumen has also upgraded its US data centre and cloud connectivity network to support Ethernet and IP services up to 400Gbps in over 70 facilities across 16 metro markets, targeting AI‑driven and latency‑sensitive enterprise workloads. 

The company couples this physical build with flexible services and conduit access products that allow customers and partners to ride its backbone rather than building their own. 

For edge‑minded operators, Lumen illustrates how to align physical, optical and service‑layer investments around AI economics.

Cisco – unifying edge compute and network

Cisco is using its scale in routing, switching and security to create integrated edge platforms that extend data centre capabilities into branch, industrial and micro‑data centre environments. 

Its new Cisco Unified Edge offering combines networking, compute, storage and security into a single, integrated appliance designed for agentic AI and real‑time applications running at the edge. 

Executives describe it as a platform rather than a server, with the intent to push data centre‑class processing power and policy control closer to data sources such as factories, retail sites and healthcare facilities.

Industry commentary suggests this approach responds directly to infrastructure bottlenecks that have stalled many AI pilots, as organisations struggle to process growing volumes of edge‑generated data within acceptable latency and cost envelopes. 

By deeply integrating compute and network functions, Cisco aims to eliminate device silos and data fragmentation, enabling edge nodes to handle full‑stack data processing and security enforcement. 

From a network expansion perspective, this architecture treats edge locations as intelligent extensions of the core, controllable through central orchestration but capable of local decision‑making. 

For operators, Cisco’s roadmap highlights the direction of travel: converged, secure and AI‑aware platforms that blur the lines between data centre and edge site.