Manufacturing leaders are under no shortage of pressure. Improve output. Reduce costs. Accelerate innovation. Respond faster to market shifts.
The good news is that a new generation of technologies — AI-powered processes, digital twins, software-defined industrial automation, autonomous robotics, virtual PLCs — offers genuine and measurable ways to address all of these challenges.
The harder news is that most manufacturers are discovering a stubborn obstacle standing between their ambitions and results: the network infrastructure on the factory floor. This isn't a technology problem in the conventional sense. The tools are mature. The business case is solid. What's holding manufacturers back is something far more foundational — and far more fixable.
The gap between ambition and execution
Across conversations with operations leaders, process engineers and automation teams, a consistent pattern emerges. They identify compelling AI use cases, secure budget, start deploying a solution in one part of their operations. And when they want to scale across plants, the project stalls — not because the technology failed, but because the underlying infrastructure wasn't designed to support it. What begins as a contained deployment quickly exposes infrastructure gaps that require unplanned investment.
Legacy industrial networks were built for a different era. They need new capabilities to enable AI pilots to scale. For example, AI-powered quality inspection systems generate continuous high-resolution image streams that need to be processed in real time. Digital twins require constant, low-latency data feeds from hundreds or thousands of endpoints simultaneously. Software-defined production systems depend on seamless, secure data flows between the shop floor, the data center and external platforms.
Each of these applications places demands on the network that most existing OT infrastructure simply cannot meet. The result is what many operations leaders experience as a ceiling — a point beyond which digital transformation initiatives cannot scale, no matter how sophisticated the applications sitting on top.
What the research shows
ARC Advisory Group, in their recent report on AI-ready manufacturing operations, frames this clearly: "Primary implementation hurdles to adoption of advanced technologies often revolve around the limitations of existing network infrastructure. Issues with network performance, flexibility, scalability, reliability, resilience, security, manageability and observability can stall or derail critical modernization initiatives."
This means that for many manufacturers, the constraint isn't the AI strategy — it's the infrastructure strategy. And infrastructure decisions, unlike software deployments, require deliberate planning, architectural thinking and organizational alignment between IT and OT teams that has historically been difficult to achieve.
Rethinking what "modern" means in an industrial context
The conversation about network modernization in manufacturing often defaults too quickly to bandwidth. More bandwidth, faster speeds — problem solved. But that framing misses the deeper shift required.
A truly modern industrial network is not simply a faster version of what came before. It is an enterprise-grade OT network, designed from the ground up to be secure, resilient, scalable and manageable — while still meeting the determinism and reliability requirements that industrial environments demand.
This means rethinking the physical infrastructure. High-wattage Power over Ethernet, capable of delivering up to 90 watts per port, is increasingly essential to power a new generation of industrial devices — high-resolution cameras, advanced sensors, edge compute nodes — without the cost and complexity of separate power infrastructure.
It also means investing in intelligent network management. As factory networks grow in scale and complexity, the gap between specialized IT networking expertise and the OT teams responsible for day-to-day operations becomes a real operational risk. AI-powered management and troubleshooting tools that allow OT technicians to monitor, diagnose and resolve network issues using plain language — without deep networking knowledge — are no longer a luxury. They are a practical necessity for maintaining uptime at scale.
Cybersecurity is a major consideration. As ARC notes, "given the scale of modern industrial networks, 'bolt-on' cybersecurity solutions are often complex, costly and ineffective. A modern approach requires a 'cyber-native' network with OT security features embedded directly into the infrastructure." Asset visibility, micro-segmentation to contain lateral threat movement, and zero-trust remote access need to be integral capabilities — not treated as a layer added afterward.
Virtualization changes the economics
One of the most consequential shifts enabled by modern network infrastructure is shop-floor virtualization — the decoupling of industrial applications from the physical hardware they traditionally ran on.
When HMIs, industrial PCs and even PLCs can run as virtualized workloads on consolidated hardware, manufacturers gain flexibility that was previously impossible. Software updates that once required physical intervention across dozens of machines can be managed centrally. New applications can be deployed without hardware procurement cycles. Hardware footprints shrink, and with them, maintenance costs.
Audi's Edge Cloud For Production (EC4P) initiative — in which the company has virtualized core production control assets — demonstrates what this can look like at scale. By treating the factory as a software-defined environment, Audi has created infrastructure that is faster to update, easier to secure and genuinely ready to absorb the next wave of AI-driven capabilities. ARC describes this approach as "a powerful example of how a forward-thinking approach to network architecture can unlock the full potential of software-defined manufacturing and AI-driven processes."
The network that makes this possible needs to be architected deliberately, with the requirements of virtualization in mind from the start.
Hear from practitioners who are doing this now
Understanding the principles is one thing. Hearing how organizations are navigating these decisions in practice is another. In a recent conversation, Arno Thijssen, senior principal engineer – Process and Automation at Keurig Dr. Pepper, described his team's approach candidly.
With a range of AI use cases already identified, they made a deliberate decision to slow down and get the foundation right first — standardizing their network infrastructure for higher bandwidth, stronger security and closer collaboration between OT and IT before scaling their AI initiatives. It is a disciplined approach that reflects a hard-won understanding of what sustainable transformation requires.
Thijssen joined Chantal Polsonetti, Vice President at ARC Advisory Group, for a webinar that covers the research findings in depth and explores the practical decisions manufacturers face when modernizing their industrial networks. The webinar is available on demand. If network modernization is on your agenda — or if it should be — it is worth an hour of your time. Watch the recording here.
The strategic implication
For manufacturing executives evaluating where to focus modernization investment, the network deserves a more prominent place in that conversation than it typically receives. Not as a back-office infrastructure decision, but as a strategic enabler—the foundation on which AI, automation and software-defined operations either scale or stall.
The manufacturers who will lead in the next decade are not simply those who deploy the most advanced AI applications. They are those who build the infrastructure capable of sustaining, scaling and securing those applications over time. The shop floor is ready for what comes next. The question is whether the network underneath it is, too.
