Why AI is redefining the industrial network
Much of the conversation around industrial AI focuses on algorithms, copilots, autonomous robots and increasingly sophisticated analytics. Yet one of the most important technologies enabling this transformation rarely receives the same attention. According to Jeremy Foster, Senior Vice President and General Manager of Cisco Compute, the industrial network has quietly evolved from an invisible utility into one of the most strategically important assets within modern manufacturing. As AI becomes embedded across production environments, the network is no longer simply moving data around the factory. It is becoming the platform on which intelligent manufacturing operates.
For decades, manufacturers viewed networking much like any other piece of plant infrastructure. As long as machines remained connected and production continued uninterrupted, there was little reason to think about it. Investment was often driven by necessity rather than strategy, with networks treated as a cost center rather than a source of competitive advantage. Foster believes that mindset is no longer sustainable.
“If you look back 30 or 40 years, networking inside manufacturing was largely plumbing,” he explains. “It moved data around and nobody really noticed what was going on unless it stopped working. With AI and autonomous operations, that whole paradigm is inverted. The network is now the common substrate across how the factory runs. Characteristics such as latency, determinism and availability directly determine what a manufacturing facility is capable of delivering.”
That shift has implications far beyond faster connectivity. AI is changing where data is processed, how operational decisions are made and how manufacturing systems continuously learn from the information they generate. Instead of supporting production in the background, the network is becoming the foundation that connects edge computing, centralized AI infrastructure, machine intelligence and operational resilience into a single architecture. Manufacturers investing in AI therefore need to think beyond deploying new models and applications. They also need to ensure the infrastructure beneath them is capable of supporting the next decade of intelligent manufacturing.
AI is redefining what the network must do
AI is forcing manufacturers to rethink not only how factories generate data, but where intelligence should be created. Traditional industrial networks were designed to transport information reliably between machines and centralized systems. AI introduces a very different set of requirements. Computer vision systems, autonomous robots, predictive maintenance applications and real-time process optimization all rely on fast, deterministic communications and the ability to process information wherever it creates the greatest operational value.
For Foster, the conversation should no longer be about whether AI runs in the cloud or at the edge. Modern manufacturing increasingly requires both. “You might have an edge computing device sitting on a manufacturing line performing computer vision, inspecting thousands of products every minute,” he explains. “The important data from that process is then sent back to a centralized location, whether that’s on-premises or in the cloud. You combine the information from multiple production lines, retrain the model, improve it and then redistribute those updated models back to the edge. That’s not something manufacturing had to think about before.”
Rather than replacing centralized computing, edge AI creates a continuous feedback loop between production assets and enterprise infrastructure. Time-critical decisions remain close to the machine, where low latency is essential, while broader operational intelligence is created by combining information from multiple production lines, factories or even global manufacturing networks. The network becomes the mechanism that enables both environments to operate as a single intelligent system.
Cisco is already seeing manufacturers build this type of architecture. One global food manufacturer is using AI-powered digital twins and real-time operational data to optimize freezing processes by automatically adjusting temperature, airflow and production speed. Running these AI workloads closer to the factory floor enables faster operational decisions that improve product quality, increase throughput, reduce waste and lower energy consumption. Supporting that deployment are Cisco AI PODs, providing the secure, low-latency compute and connectivity needed to process these workloads close to production while integrating them with enterprise AI infrastructure.
Data, not hardware, creates competitive advantage
As manufacturers accelerate AI adoption, it is easy to assume competitive advantage will come from deploying more powerful hardware or the latest AI models. Foster believes that view fundamentally misunderstands where long-term value will be created. While access to GPUs, edge computing and AI software is becoming increasingly widespread, the real differentiator is how effectively manufacturers can capture, connect and exploit their own operational data.
“If everyone has access to GPUs, buying one and putting it next to your robot doesn’t make you better than your competition,” Foster says. “What makes you better is how you extract more value from your own data and use it to make your manufacturing operations more efficient. AI has to help you make sense of all that information at a speed you could never achieve with manual processes.”
That places new demands on the industrial network. Rather than simply transporting information, it must securely connect machines, production lines, edge infrastructure and enterprise systems while ensuring data can move efficiently to wherever it creates the greatest value. As AI models continue to evolve, manufacturers also need the ability to retrain, update and redeploy them across multiple sites without disrupting production, creating an infrastructure that continuously improves alongside manufacturing operations.
Cisco’s Unified Edge platform was developed with that operational challenge in mind. Rather than treating networking, compute, storage and management as separate technologies, it brings them together into a single platform that can be deployed and managed consistently across hundreds or even thousands of distributed edge locations. Foster argues that managing AI infrastructure at scale is becoming just as important as deploying it in the first place, particularly for manufacturers operating multiple plants.
That approach is already finding practical applications. One of North America’s largest dairy processors is standardizing AI-ready infrastructure across its production facilities, creating the foundation for machine vision, automated quality control and future edge AI applications while enabling remote collaboration and faster troubleshooting across multiple manufacturing sites.
AI is making resilience a design requirement
As factories become more intelligent, they also become more interconnected. As production assets, AI models and operational data become increasingly connected, the traditional approach of isolating systems behind network boundaries becomes far more difficult to sustain. Information that was once confined to individual production cells now needs to move securely between machines, edge platforms, enterprise applications and cloud infrastructure, creating a much larger operational ecosystem.
For Foster, that makes cybersecurity and resilience fundamental design principles rather than capabilities that can be added later. “You can’t just put things on an island anymore,” he explains. “The data being generated across the factory has become valuable, and it’s going to be accessed by lots of different systems. As you link all these previously separate environments together, the attack surface inevitably grows. That’s why security, observability and modern networking have to be designed together from the start.”
That philosophy is reflected throughout Cisco’s AI infrastructure. Rather than treating security as a separate layer, capabilities such as Zero Trust, observability and policy management are embedded within its reference architectures, enabling manufacturers to modernize IT and OT environments without compromising operational resilience. The company’s design guidance also recognizes that infrastructure will need to evolve continuously as AI workloads, production requirements and cyber threats change over time.
For manufacturers, futureproofing therefore extends beyond deploying enough compute capacity to support today’s AI applications. It requires building an infrastructure that can accommodate new models, connect additional assets, manage thousands of distributed edge devices and maintain consistent security across every production facility. Foster argues that manufacturers should think about these investments over a five- to ten-year horizon. The infrastructure decisions being made today will determine not only how quickly AI can be deployed, but also how effectively factories continue to evolve as intelligent, connected operations over the decade ahead.
Infrastructure becomes competitive advantage
The conversation around industrial AI often focuses on what the technology can do. Foster’s message is that none of those innovations can deliver their full potential without the infrastructure capable of supporting them. As manufacturers continue to connect more assets, deploy more AI at the edge and rely on real-time operational intelligence, the industrial network is becoming the platform that enables every one of those capabilities to work together.
“Let’s not treat the network as a necessary evil or a means to an end,” he concludes. “Let’s understand that this is now going to be the thing in a modern manufacturing facility that’s going to enable you to operate your business end to end.”
For manufacturers planning the next phase of their digital transformation, that may be the most important lesson of all. Competitive advantage will depend not simply on adopting AI, but on building an infrastructure that allows intelligence to scale securely, reliably and continuously across the entire enterprise.
