By: Jayson Kordich, Senior Product Marketing Manager, Zayo
A year ago, we launched DynamicLink with a simple premise: enterprise connectivity should be easier to consume and control without sacrificing the performance, reach, and reliability of the network underneath it. Twelve months in, that premise has held up better than I expected: the future of networking can’t just be about faster connectivity; it must also be about putting enterprises in control when and where they want it.
The network was built for a world that no longer exists. For decades, its primary job was straightforward: connect people to applications, locations to locations, and businesses to the internet. That connectivity remains essential, but how enterprises consume and operate it has fundamentally changed – from the proliferation of MPLS to the Cloud era and beyond.
Now? The Intelligence Era asks more of the network. Data may live in one location, compute and AI services in another, and the people or systems that need the result somewhere else entirely. Applications scale dynamically. AI workloads can appear quickly and move enormous amounts of data. Cloud architectures expand across regions. More intelligence is moving toward the edge. Scale up, down, across. Data everywhere, all at once, all the time. None of this is a surprise, but it is new.
As traffic patterns change, the operating model around the network must change with it. That’s the gap DynamicLink was built to close, and it’s worth revisiting why, now that we have a year of real customer use behind us.
The Gap is Widening
AI is accelerating a shift that was already underway.
For years, bandwidth demand grew in relatively predictable increments. Organizations added capacity as users, applications, and cloud adoption expanded. AI has made those requirements considerably less predictable, increasing both the amount of capacity organizations need and how quickly they may need to change it.
Zayo’s own purchasing data shows the scale of the shift. Between 2024 and 2025, demand for long haul dark fiber doubled as large customers connected critical AI data centers. In leading AI-driven metros, Zayo has seen dark fiber demand reach as much as 20 times that of other markets, while 400G continues to represent a significant share of wavelength capacity sold. You can find more of these insights in Zayo’s 2026 Bandwidth Report.
Those trends reinforce something important: the physical network matters more than ever.
Dark fiber, wavelengths, Ethernet, internet, and cloud connectivity provide the capacity, performance, and reach that modern applications depend on. The challenge isn’t that these foundational services are becoming obsolete. It’s that enterprises increasingly need to use, combine, and change them in ways that traditional provisioning models weren’t designed to support.
That’s creating a divide between organizations that can adapt their networks as requirements change and those still operating through long planning and provisioning cycles. Sure, more bandwidth is necessary. But capacity without agility is just a bigger network with the same limitations.
The Variable That Matters Most is Speed of Change
It’s tempting to frame AI networking primarily as a capacity challenge: add bigger pipes and keep moving. Capacity only addresses one dimension of the problem.
An enterprise may also need to connect a dataset in one cloud to an AI service in another, deliver the output to an enterprise location, increase capacity for a new workload, or establish connectivity for a new site. Increasingly, those requirements emerge on business timelines rather than traditional telecom timelines.
How quickly the network can change now directly affects how quickly the business can move and compete. theCUBE Research found that 93% of organizations now consider the network more important to achieving business goals than it was just two years ago. As the network becomes more central to applications, cloud, and AI strategies, enterprises need an operating model that can respond in kind.
The standard now looks less like traditional telecom provisioning and more like cloud infrastructure: available when needed, configurable through software, observable in real time, and accessible to automation. That’s the model behind DynamicLink, and it’s what my colleague Ted Wagner, who leads the DynamicLink product, and I spent most of our Sept. 30 webinar unpacking with customers.
What a Year of Customer Use Confirmed
“Being able to manage your connectivity the moment you need it is huge.”
DynamicLink is Zayo’s approach to Network-as-a-Service — redefining what NaaS can actually do when it’s a core part of networking, instead of disparate, temporary, or limited “on demand” solutions.
It provides a software-based layer for establishing and modifying network services, rather than treating every new requirement as a separate provisioning exercise. Enterprises can use ports at the locations that matter and configure services such as dedicated internet access, Ethernet, and cloud connectivity through a portal or APIs.
A year of customer deployments reinforced three lessons about what enterprises actually need from a modern network.
- One network, one solution, more possibilities. DynamicLink brings Layer 2 and Layer 3 services – including security and observability – together under one carrier-grade, SLA-backed platform. Instead of designing around the limitations of disconnected providers and services, enterprises gain the freedom to build network architectures around their business needs, with greater flexibility to connect, combine, and evolve services as those needs change.
The proof: I’ve seen customers go from experimentation to super-user in the span of a few months – when that “lightbulb moment” occurs and they see the freedom and control DynamicLink unlocks, that’s when the magic happens. “You mean I can…?” Yep. You can!
- Less network management creates more business value. DynamicLink simplifies service management by putting control of the network in one place. Through the portal, APIs, DashAI, or AI agents connected via MCP, teams can eliminate repetitive manual tasks, reduce time lost to provisioning and change orders, and manage services with far greater efficiency. The payoff goes beyond time savings: leaner operations, faster execution, and more resources focused on strategic priorities rather than keeping the network running.
The proof: A cheeky email from a software customer of ours with a picture of a pot of coffee – they were referencing a video we did on DynamicLink earlier this year. “Put on a pot of coffee, and by the time it’s done brewing…” – this customer shared hours of time saved per month because with DynamicLink, they weren’t on the hook coordinating providers, submitting change requests, and waiting for services to be provisioned.
- An intelligent network that goes wherever business takes you. Modern enterprises shouldn’t have to design their networks around a limited set of available locations. DynamicLink combines rapid, software-defined provisioning with Zayo’s expansive infrastructure and off-net connectivity options to bring dynamic networking capabilities to enterprise locations across the U.S. and Canada, not just a few data centers. With real-time visibility, intelligent automation, and the ability to activate or adapt services in seconds where connectivity is established, the network becomes a responsive resource that keeps pace with the business, not a constraint holding it back.
The proof: Working with a healthcare company finishing up an acquisition, they needed consistent connectivity across all their locations without being confined to a handful of interconnection hubs. Together, we configured a DynamicLink architecture that supported these locations using a multipoint layer 2 WAN, including direct links to AWS and Azure.
We’ve also started shifting how enterprises pay for this. Most customers began with a fixed monthly subscription that includes connectivity, protection, observability, and operations capabilities, with usage-based billing and event-driven automation planned next to support more flexible consumption models as requirements change.
DynamicLink has become more than a faster way to order connectivity. It’s becoming the intelligent control layer that turns network infrastructure into something enterprises can command, adapt, and automate. As we look toward the next decade of networking, one thing has become quite clear to me: 2026 set a new standard for what networks should be capable of.
Agentic Networking: What It Means for the Network to Act
Most AI integrations in networking stop at the interface. They let someone ask a question and get an answer back. Zayo has gone further, launching Agentic Networking, the first production capability from a network service provider that lets approved AI agents take authorized action on live WAN services rather than simply query them.
Agentic Networking builds on the same foundation DynamicLink established over the past year. Enterprises already use the platform to understand network context and execute self-service changes through a portal or APIs. Agentic Networking extends that foundation with a purpose-built networking skill and a Model Context Protocol server, an emerging standard that gives AI agents a common language for understanding and interacting with a platform’s capabilities.
That extension matters because enterprises aren’t limited to a Zayo-built agent. A customer with its own AI agent can connect it to the network and put it to work within their defined boundaries. Organizations decide what information, tools, and actions are available to their agents, and approval and governance stay in place throughout.
The distinction is worth stating plainly. Earlier generations of network AI (which to be clear was not very long ago) were largely read-only, trained on static documentation, and limited to answering questions. That model works for a support chatbot. It doesn’t work for a network engineer who needs an agent to provision a connection, audit a firewall rule, or troubleshoot a performance issue.
With Agentic Networking, AI isn’t just telling you how to modify your network – it’s actually making that change happen, within the permissions and approvals you’ve defined.
During the Sept. 30 webinar, Ted and I walked through what that looks like in practice. An AI-assisted workflow might interpret a request, determine the required configuration, present the proposed change, and allow an operator to approve it before anything touches the live network. The operating loop is straightforward: observe, understand, recommend, act, and verify.
The goal is straightforward: reduce the manual work required to turn intent into action while keeping people in the decisions that matter. Some operations will stay human-controlled. Some will be automated. Others may be managed on the customer’s behalf. Enterprises may choose to extend greater authority to trusted automation and AI agents over time, but the network needs to support all of those models now, not just the fully automated end state.
Cloud Strategy No Longer Has to Begin at The Data Center
The same shift in thinking applies to cloud architecture.
For years, private cloud connectivity was commonly anchored around the data center. Enterprise traffic traveled back to an aggregation point or cloud on-ramp before reaching AWS, Azure, or Google Cloud. That architecture remains valuable, particularly where aggregation, colocation, and interconnection matter. It simply doesn’t have to be the only model.
As AI shifts from centralized training toward more distributed inference, data increasingly needs to move between clouds, enterprise locations, edge environments, and users. The network needs to make cloud connectivity available wherever the business requires it.
DynamicLink extends that software-defined connectivity model beyond a small set of interconnection facilities. A headquarters, campus, manufacturing environment, data center, or other connected enterprise location can increasingly participate in the broader cloud architecture. The question becomes less about how every location gets back to the data center and more about where each location should connect based on the workload.
This doesn’t diminish the role of data centers. It gives enterprises more architectural freedom, letting cloud strategy follow the data, compute, and users rather than being dictated by the location of an interconnection facility.
Intelligence Still Depends on Infrastructure
None of this makes the physical network less important. Quite the opposite.
AI training is driving demand for massive capacity between data centers. Inference is increasing the importance of metro connectivity and bringing applications closer to users. Distributed cloud strategies require reliable connections between more environments. Real-time applications make latency, routing, and performance increasingly consequential.
DynamicLink sits above that foundation rather than replacing it. Zayo’s Dark Fiber and Wavelength infrastructure provides the capacity required for the largest data movements. Ethernet, internet, and cloud connectivity connect enterprise environments to the applications and infrastructure they depend on. Zayo’s expanding long haul and metro footprint provides the physical reach underneath those services.
Infrastructure provides the foundation. Software puts it under your control. Intelligence makes it adaptive. A year into DynamicLink, I’m convinced the next network needs all three, and that no single one works as a substitute for the other two.
What the Network Lets Enterprises Do Next
The organizations pulling ahead in the Intelligence Era won’t simply buy more bandwidth. They’ll rethink the network as a strategic resource that can respond as quickly as the infrastructure and applications around it.
That doesn’t mean replacing what works. It means getting more out of it. Connectivity becomes easier to consume. Capacity becomes easier to scale. Network behavior becomes more visible. Cloud architecture becomes more flexible. AI and automation are taking on more operational tasks, while people stay in control.
The underlying network remains the foundation. Last October, we set out to make connectivity more dynamic. Today, we’re helping enterprises rethink what their networks can actually do – and that they should expect more from their network. The case for dynamic connectivity isn’t theoretical anymore.
The good news? We’re just getting started.
Put DynamicLink to the Test
If your team is still waiting weeks for connectivity changes the business needs now, the delay is more than an operational inconvenience. It limits how quickly the organization can respond.
The most useful way to understand a different operating model is to experience it. Talk with a Zayo connectivity expert and bring us a connectivity challenge your team is facing today. We’ll explore how a more dynamic approach could fit your architecture and give you an opportunity to try DynamicLink for yourself.