LAS VEGAS — The technology industry has a funny habit. Once a technology becomes widely adopted, we stop talking about it. It fades into the background and becomes part of the scenery. We call it infrastructure. Sometimes we call it plumbing.
That label often obscures how much value can be created by controlling those layers.
For decades, some of the biggest fortunes in technology were built not on consumer applications or flashy user experiences, but on infrastructure that most people never thought about. Cisco built an empire on networking. Intel did the same with processors. Both companies became dominant by owning layers that every enterprise depended on, even if few executives woke up in the morning excited to discuss routers or CPUs.
The irony is that infrastructure usually becomes most visible when it starts breaking or when the workload changes.
Artificial intelligence has changed the workload.
The first phase of the AI boom focused almost entirely on semiconductors. NVIDIA became the face of that transformation, and deservedly so. The company recognized earlier than most that AI represented more than another application running on existing hardware. AI required a different approach to computing. The combination of GPUs, accelerated computing and the software ecosystem built around CUDA fundamentally altered the economics of the semiconductor industry.
NVIDIA’s rise is often discussed as a story about market share. It is more accurately a story about market definition. The company succeeded because it recognized that AI would change what mattered most in computing and positioned itself around that shift before its competitors did.
That distinction matters because semiconductors are unlikely to be the only layer transformed by AI.
The industry spent the last several years focused on compute because compute was the most obvious constraint. Access to GPUs became a strategic concern. Data center capacity became a strategic concern. Even power availability became a strategic concern. The conversation around AI often sounded as though every problem eventually led back to the same answer: more compute.
As organizations begin moving AI workloads into production, however, the conversation is broadening. Models rarely operate in isolation. They need access to data, applications, workflows and users. Increasingly, they also need to interact with other models and autonomous agents. The challenge becomes less about generating intelligence and more about moving information, enforcing policy and maintaining visibility across increasingly distributed environments.
That shift brings the network back into focus.
Not because networking suddenly became exciting again, but because AI systems place new demands on how information moves through organizations. Large AI deployments require enormous amounts of data movement. Inference increasingly occurs across clouds, data centers, campuses and edge locations. Security policies must travel with workloads. Operations teams need visibility into environments that are becoming more dynamic and more interconnected at the same time.
The network sits in the middle of all of it.
That reality provides an important backdrop for Cisco Live this year. Cisco is not arriving in Las Vegas simply to announce another collection of products with AI features attached. The company is making a larger argument about where enterprise AI is heading and what organizations will need to operate it successfully.
At the center of that argument is the idea that AI infrastructure requires a control plane.
The phrase “operating system for AI infrastructure” captures what Cisco appears to be pursuing. The company is not trying to build foundation models. It is not trying to become the next Nvidia. Instead, Cisco is positioning itself as the layer responsible for connecting, securing, observing and operating increasingly complex AI environments.
Viewed through that lens, many of Cisco’s major investments over the last several years begin to fit together in ways that may not have been obvious when they were announced.
The acquisition of Splunk expanded Cisco’s ability to observe and understand what is happening across complex environments. Security investments addressed the reality that every new AI capability introduces new governance and risk-management challenges. Silicon One and optical networking initiatives acknowledged that moving data efficiently is becoming just as important as processing it. Automation efforts recognized that infrastructure operating at an AI scale cannot be managed through traditional approaches.
Individually, each of those initiatives can be explained on its own merits. Collectively, they resemble a company assembling the components of an operational layer for AI infrastructure.
Whether that vision becomes reality is one of the most important questions facing Cisco.
Technology history is filled with companies that dominated one era and struggled in the next. Success creates its own challenges. The assumptions that make a company successful often become difficult to abandon when markets change. Organizations frequently spend too much time defending the categories they already lead instead of adapting to the categories emerging around them.
That is one reason the comparison between Cisco and Intel is so compelling.
Intel spent years operating from a position of strength in a market it helped define. When AI began reshaping the semiconductor landscape, the company found itself responding to changes that were already underway. NVIDIA, by contrast, benefited from helping define the new market around AI workloads.
Cisco appears determined not to find itself in a similar position.
Rather than defending a traditional definition of networking, the company is attempting to expand the definition of what networking means in an AI-driven world. Networking, in Cisco’s telling, increasingly includes security, observability, automation, policy enforcement and operational control. Whether customers ultimately agree with that definition remains to be seen, but the strategic logic behind it is difficult to miss.
The timing is also significant.
For years, much of the AI conversation has centered on model builders and chip suppliers. Those companies remain critically important, but enterprises eventually have to move beyond experimentation. They must integrate AI into business processes, secure it, govern it, monitor it and operate it at scale. The closer AI moves toward production, the more infrastructure questions begin to matter.
That transition may ultimately benefit companies whose expertise has always been infrastructure.
As we spend the week talking with Cisco executives, engineers, partners and customers, that will be the lens through which we evaluate the announcements and demonstrations coming from the event. The most important question is not whether Cisco can add AI capabilities to its portfolio. Every major technology vendor is doing that.
The more important question is whether Cisco can convince customers that operating AI infrastructure is becoming a distinct challenge of its own and that the company is uniquely positioned to solve it.
That is a much larger ambition than selling networking equipment. It is an attempt to become the operational layer through which enterprise AI is deployed, secured and managed.
Cisco has spent decades building businesses around infrastructure that others viewed as plumbing. AI is creating conditions where that plumbing is becoming strategic again.
The company believes it knows this playbook well. The question hanging over Cisco Live is whether the rest of the market agrees.
