In this exclusive op-ed for The Manufacturer, Paul Williamson, SVP & GM, IoT Line of Business, Arm, looks at how the convergence of IoT and edge AI is revolutionising manufacturing by enabling real-time, autonomous decision-making directly on the factory floor, driving greater efficiency, resilience and agility while reducing reliance on cloud infrastructure.
From vision-guided robots to predictive maintenance systems that act before failures happen, the convergence of the Internet of Things (IoT) and edge artificial intelligence (edge AI) is transforming manufacturing from the inside out. As industrial environments grow more digitised and decentralised, the ability to process data and make decisions in real time – right at the edge – has become a defining feature of modern factories. Edge AI refers to artificial intelligence that runs directly on devices like sensors, cameras, and controllers, on or near the production floor, rather than relying on distant cloud data centres.
While much of the early focus in Industry 4.0 revolved around connectivity and cloud analytics, a shift is well underway: intelligence is moving from centralised data centres to the edge of operations. In today’s high-speed, high-stakes industrial environments, milliseconds matter and unplanned downtime can cause significant disruption. That’s why manufacturers are turning to edge AI to power the next wave of transformation – smarter, safer, and more sustainable systems that work independently, provide faster responses and ensure robust, local autonomy, even in the most remote or power-constrained conditions.
A sector on the leading edge of AI adoption
AI has quickly established its value in the industrial sector – and for good reason. The economic impact of unplanned downtime is substantial, costing Fortune Global 500 companies an estimated £1.1tn ($1.5tn) annually, or more than 11% of their turnover. But predictive maintenance, in turn, can cut downtime by 75%.
Given the numbers, it’s no surprise that the recent Arm AI Readiness Index report shows 87% of manufacturing business leaders are already using AI applications to address these issues. That’s more than any other sector surveyed in the report, and places manufacturing at the forefront of AI readiness. Nearly half deploy AI specifically in manufacturing processes such as quality control and process optimisation. And perhaps most notably, one in three manufacturing organisations actively use IoT and sensor data to power these AI initiatives – almost double the rate seen in other sectors.
This puts manufacturers in a unique position: they’re not just collecting data, they’re increasingly turning it into actionable intelligence on the factory floor. From real-time defect detection to adaptive robotics, the integration of edge AI is enabling faster decision-making and reduced reliance on cloud infrastructure, ushering in a more agile era for industrial operations.
Where latency meets the bottom line
Edge computing has emerged as a linchpin for smart manufacturing strategies. By enabling AI models to run directly on embedded devices such as smart cameras, sensors, and microcontrollers, factories can respond to critical signals in real time without waiting for data to travel to the cloud and back. This reduces latency, conserves bandwidth, and enhances system resilience.
Take predictive maintenance: sensors monitor vibrations, heat, and sound to detect early signs of mechanical failure. When these models run at the edge, systems can autonomously alert operators or trigger preemptive shutdowns – before damage occurs. Not only does this minimise unplanned downtime, but it also improves equipment longevity and total cost of ownership. This capability is especially valuable in remote facilities or environments where connectivity is intermittent.
From quality control to cobots: AI in action
Across the manufacturing landscape, edge AI is taking on more complex, mission-critical roles, such as computer vision. Smart cameras equipped with embedded processors can now perform real-time image analysis to detect anomalies on fast-moving production lines. Rather than sending images to a central server for inspection, these systems can instantly flag defects or trigger automated responses, drastically improving quality assurance and throughput.
Another standout use case is collaborative robotics. Known as ‘cobots’, these robots are increasingly equipped with AI that enables them to work safely and adaptively alongside humans. They can automate risky or repetitive tasks while enabling workers to take on higher-value roles. This shift supports long-term workforce sustainability and aligns with broader organisational goals around digital transformation and employee upskilling.
Using edge-powered sensors and cameras, cobots can detect obstacles, adjust their grip strength, or change task sequences based on environmental cues. This enables more flexible automation–ideal for high-mix, low-volume manufacturing where conditions change frequently.
A case in point, R2C2, a robotics software company, is enabling robot dogs to autonomously inspect train cars in Hong Kong, checking thousands of safety points using computer vision –achieving over 99% inspection accuracy while reducing inspection time from hours to minutes.
A shift toward decentralised and resilient operations
Global supply chain disruptions over the past few years have prompted manufacturers to rethink centralised production models. As a result, there’s been a growing movement toward decentralisation–smaller, distributed facilities that can operate semi-autonomously. Here again, edge AI plays a critical role.
By placing intelligence at the edge, manufacturers empower local operations to make real-time decisions even without continuous cloud connectivity. This distributed architecture supports operational continuity, faster response times, and improved resilience in the face of logistical or network disruptions.
Power and infrastructure: the next challenge
Despite strong adoption, edge AI still faces hurdles – especially around infrastructure and energy demands. Only 29% of manufacturers report they can automatically scale compute or storage resources to meet AI workloads, and just 23% have dedicated power infrastructure in place, according to the AI Readiness Index report. These constraints make energy-efficient edge AI solutions critical for continued growth.
To support widespread deployment, edge devices must be able to run advanced models within tight power and thermal budgets. This is why manufacturers are increasingly prioritising ultra-low-power processors and efficient model architectures that can perform reliably in rugged, resource-constrained environments – from offshore rigs to compact production lines.
Security and compliance from the ground up
As more intelligence shifts to the edge, cyber security is a top concern. Nearly half of manufacturing leaders worry about data privacy breaches from model extraction, and over a third cite integration with existing security systems as a major challenge.
Processing sensitive data locally, rather than transmitting it to the cloud, can help mitigate these risks. Edge AI reduces the number of potential attack vectors, supports compliance in regulated industries, and enhances data sovereignty – a key consideration for globally distributed manufacturers.
Welcome to manufacturing’s AI era
With AI budgets expected to increase across manufacturing organisations over the next three years, according to the Readiness Index report, the momentum behind edge-enabled Industry 4.0 continues to build. But to fully capitalise on this transformation, manufacturers must prioritise integration-ready platforms that simplify deployment, scale efficiently, and deliver strong compute performance within tight energy constraints.
For senior executives navigating digital transformation, successful edge AI adoption requires a strategic, phased approach. Organisations should begin by conducting a comprehensive audit of infrastructure readiness, evaluating whether current power, compute, and connectivity setups can adequately support distributed AI workloads at the edge. Rather than attempting enterprise-wide rollouts, executives should target high-ROI pilot projects that focus on use cases with measurable impact, such as visual inspection, predictive maintenance, or energy optimisation. Equally critical, is fostering collaboration between IT and operational technology (OT) teams to ensure cross-functional alignment on data governance, security protocols, and systems integration – ultimately maximising the operational value of AI across the entire plant floor.
The future of smart manufacturing isn’t just connected – it’s intelligent, autonomous, and resilient. And it’s being built not in faraway data centres, but right at the edge of operations.
About the author
Paul Williamson is Senior Vice President and General Manager, IoT Line of Business, at Arm.
Paul leads the IoT line of business at Arm where he and his team are working alongside the Arm ecosystem to bring compute to a diverse set of applications, unlocking value with intelligence and automation.
Previously at Arm, Paul has led the client line of business, shaping the future of consumer products, as well as the security, IoT, and wireless businesses.
Prior to joining Arm, Paul led the low-power wireless division of CSR, a fabless semiconductor business (now part of Qualcomm). Paul started his career in engineering consultancy, working with leading global brands to develop innovative products and services. He holds an MEng from Durham University.
For more articles like this, visit our Industrial Data & AI channel