Network-as-a-service in practice: Lumen, AT&T, Verizon and Megaport, and the architecture they make possible

What NaaS means now, and what the industry is saying

The industry definition comes from Mplify Alliance, the association formerly known as MEF, which took the new name in June 2025. Its NaaS Industry Blueprint, updated in October 2024, defines network-as-a-service as integrating on-demand connectivity, application assurance, cybersecurity and multi-cloud networking within an automated ecosystem. In plain terms: the carrier exposes its network through a portal and APIs, you order and change services in minutes, and you pay for what you use.

The analyst view is that this is where enterprise networking is heading. ABI Research forecast in January 2023 that by 2030 over 90 percent of enterprises would consume at least a quarter of their network services on a usage basis. The same release warned that uncertainty about cost keeps the market early. Omdia lists demand for NaaS functionality among its 2026 trends to watch in enterprise connectivity. The same software-defined platforms that run SD-WAN and SASE are what make it possible: once the network is programmable, consumption pricing follows.

The caution is also worth hearing. NaaS changes how you buy, not whether the network matters. The circuits still have to reach your buildings, the support still has to answer, and the contract still has to say what happens when it does not.

Lumen NaaS: switched Ethernet and internet, on demand

Lumen’s model starts with a multiservice Lumen Fabric Port at your location or data center, activated once by a technician, and the services ride on it. Lumen Intelligent Internet is internet with elastic capacity you activate and scale in minutes, from 1 Mbps to 100 Gbps. Ethernet Fabric Connect creates private connections between sites, data centers and clouds, on-net and off-net, with bandwidth you scale in minutes as the need changes, and Cloud Interconnect adds automated private Layer 3 connections to the cloud providers. Everything is provisioned and managed through the Lumen Connect portal or by API, billed on consumption, with a minimum one-year term on the offer. Lumen’s threat intelligence team, Black Lotus Labs, protects the network upstream.

Where we use it: a seasonal site that needs real bandwidth for part of the year; a migration window that needs a temporary private path to a cloud region; a disaster-recovery test that needs capacity for a week; and as the fast way to light a new building while a long-term design is settled.

AT&T Switched Ethernet on Demand

AT&T’s version, Switched Ethernet on Demand, gives each location a port with a standard Ethernet interface of 100 Mbps, 1 Gbps, 10 Gbps or 100 Gbps. From a web dashboard you set the bandwidth on each port and the logical connections between ports, any-to-any or hub-and-spoke, with changes applied in near real time. You pay only for the bandwidth you use, so a site with a 1 Gbps port can run far below that outside its busy season and turn it up when needed. A site that outgrows its port moves to a higher-speed port.

Where we use it: a multi-site WAN where bandwidth needs vary by site and by season, where the customer wants one private Ethernet fabric across its locations and the control to resize it without a change order.

Verizon NaaS

Verizon positions NaaS as a family rather than a single product: a virtualized, programmable network bought as a service, with managed SD-WAN, managed network services, virtual network services and connectivity under one umbrella. The pitch is rapid scaling, centralized control and visibility of network resources, and a programmable foundation that evolves as the technology does.

Where we use it: organizations that want network, SD-WAN and connectivity managed as one service with one operating model, and that value that over hands-on control of each circuit.

Megaport: bring your own connectivity to the fabric

Megaport is a different model. Rather than delivering circuits to your buildings, it operates a software-defined network with more than 1,200 enabled locations in more than 30 countries, and you bring your own connectivity to one of its points of presence. The pieces: a Port, your physical entry onto the fabric in a data center, at 1, 10 or 100 Gbps depending on location; Virtual Cross Connects (VXCs), private point-to-point Layer 2 connections from that port to a destination on the fabric, sized from 1 Mbps to 100 Gbps and resized in minutes; the Megaport Cloud Router (MCR), a managed virtual Layer 3 router that peers privately between public clouds and providers without equipment of your own; and Megaport Virtual Edge (MVE), which runs SD-WAN, virtual routers from the major vendors and firewall-as-a-service on the fabric for traffic to cloud and branch locations. Destinations include the major cloud providers’ on-ramps, SaaS and service providers in the Megaport Marketplace, your own other locations, Megaport Internet and internet exchanges.

Where we use it: any design where the data lives in colocation and in more than one cloud. A private cloud in a colocation facility gets a Port; VXCs reach each cloud region in use; an MCR routes between clouds without hair-pinning through the private environment; MVE brings branches onto the same fabric; and a carrier NaaS or a conventional circuit carries the sites into the facility.

Architecting for flexibility, no matter where the data lives

1. Decide the entry points. Which sites get carrier NaaS ports, which data centers get fabric ports, and where the two meet.
2. Connect by flow, not by diagram. Each application flow gets a connection sized for it: committed bandwidth for the steady flows, on-demand for migrations, tests and seasons.
3. Keep the clouds private. Private connections to each cloud region, routed between clouds on the fabric, so traffic stays off the public internet and egress is billed under each cloud provider’s private-connection rates rather than its internet rates. Check the current rate card before counting a saving.
4. Design the diversity. Two ports, two paths, two entrances for anything that matters; on-demand does not mean single-threaded.
5. Write the exit. On-demand services are easy to start and should be easy to stop; confirm the terms say so, and that monitoring and support match what the business needs on day two.

Done this way, a new cloud region, a new SaaS platform or a new site becomes a connection ordered from a portal rather than a procurement project, and the organization can move workloads between private and public environments as costs and requirements change.

Choosing between them

Carrier NaaS from Lumen, AT&T or Verizon fits when the carrier is on-net at your sites and you want to buy connectivity itself on demand. Megaport fits when your data is already in colocation facilities and clouds and you want to interconnect them on demand. Most multi-site organizations with any cloud footprint end up with both: carrier services to the buildings, the fabric between the data centers and the clouds. Because we work with all of these providers, we can qualify which ones are on-net and which model fits each part of the design, then order and manage them as one network.