What is edge AI, and why is it used on the factory floor?

Processing Data Close to Where It’s Created

Edge AI refers to running machine learning models directly on local computing devices situated near the equipment or sensors generating data, rather than sending all raw data over a network to a centralized data center or cloud platform for processing. In a manufacturing context, this typically means small, ruggedized computing devices installed on or near production equipment, capable of running AI inference — the process of applying a trained model to new data — right where the data originates.

This is a meaningful architectural choice, not just a technical detail. It reflects a deliberate trade-off between the computing power available in centralized cloud infrastructure and the speed and reliability benefits of processing data locally.

Why Speed Matters So Much in Manufacturing

Many manufacturing applications are highly time-sensitive. A computer vision system inspecting products on a fast-moving production line needs to make a pass/fail decision within a fraction of a second, fast enough to keep pace with the line without causing a bottleneck. If that decision required sending each image to a distant cloud server, waiting for analysis, and receiving a response back, the round-trip delay — even if only a fraction of a second — could be enough to make the system impractical for a line moving at production speed. Edge AI eliminates this round trip by performing the analysis locally, delivering a decision with minimal delay.

Beyond speed, edge AI also reduces the sheer volume of raw data that needs to travel across a factory’s network. Continuous sensor streams and especially high-resolution video or image data can be enormous in volume; processing this data locally and sending only the relevant, distilled results — like a defect flag or an anomaly score — significantly reduces the burden on network infrastructure and the associated costs of transmitting and storing that raw data centrally.

Resilience and Reliability Benefits

Another practical advantage of edge AI is that certain functions can continue operating even during a temporary disruption to internet or broader network connectivity, since the processing happens locally rather than depending on an active connection to a remote server. For manufacturing environments where any downtime is costly, this added resilience can be a meaningful benefit, ensuring that critical monitoring or inspection functions don’t simply stop working if a network connection drops.

The Trade-Offs of Local Processing

Edge AI isn’t without its own limitations. Local computing devices generally have less processing power and memory than centralized cloud infrastructure, which can constrain the size and complexity of the AI models they can run effectively. There’s also added operational complexity in managing, updating, and monitoring AI models deployed across potentially many distributed edge devices throughout a facility, compared to maintaining a single centralized system. Because of these trade-offs, many manufacturers use a hybrid approach: handling immediate, time-sensitive decisions at the edge, while still sending data to the cloud for deeper, more computationally intensive analysis that doesn’t require an instant response, such as long-term trend analysis or model retraining.

Bottom Line

Edge AI runs machine learning models directly on local devices near factory equipment rather than relying solely on centralized cloud processing, primarily to achieve the low latency and reduced network load that time-sensitive manufacturing applications require. It’s typically used alongside, rather than as a full replacement for, cloud-based analytics, combining fast local decisions with deeper centralized analysis.

Go deeper

• Is Local AI More Private Than Cloud-Based AI?
• What Are the Tradeoffs Between Running AI in the Cloud vs. Locally?
• Which Businesses Benefit Most From Local AI Deployment?