Industrial AIoT moves deeper into edge operations

AIoT adoption is accelerating across industrial sectors as mature IoT platforms, edge hardware and clearer regulation make deployments more practical. Software AG’s Cumulocity has launched its ‘Helios’ AI engine, adding generative AI features that create high-fidelity synthetic data for training predictive maintenance models when historical datasets are incomplete.

Platform-based tools are lowering adoption barriers for heavy manufacturing, logistics and utilities through pre-built modules and low-code environments. Stadler Rail UK has started a pilot using vibration and temperature data from trackside equipment to predict rail switch failures with over 95% accuracy, illustrating the shift from reactive maintenance toward uptime and safety gains.

AI processing is also moving closer to machines, with NVIDIA’s ‘Jetson Thor’, Google’s ‘Coral TPU Gen 3’ and the ‘Turing-5’ edge processor supporting low-latency industrial uses. The ‘Turing-5’ chip promises a threefold performance increase for computer vision models used in anomaly detection, while edge AI cameras in a bottling plant can inspect 100 bottles per second and trigger automated removal of flawed units.

Generative AI is finding operational roles in shift reports, factory simulations and natural-language access to sensor networks, including an E.ON pilot in Germany. The UK Government’s Department for Science, Innovation and Technology has issued updated draft guidance emphasizing auditability and verifiability, while Axonflow secured a £150 million Series C funding round for its automotive battery defect-detection platform.