The cloud model — centralise compute, send all data to a regional data centre, process and return results — works brilliantly for most applications. But there is a class of applications where sending data to the cloud first creates unacceptable latency, consumes prohibitive bandwidth, or violates data sovereignty requirements. For these applications, edge computing is not an alternative to cloud — it is the completion of the cloud architecture.
Edge computing moves processing closer to where data is generated: factory floors, retail stores, hospital wards, vehicles, and smart infrastructure. A quality inspection camera on a manufacturing line that processes 120 frames per second cannot send all of that data to a cloud region and wait for a classification result. The decision — pass or fail — must be made in milliseconds, on the device.
The architecture of an edge computing deployment has three layers. At the far edge — the device layer — lightweight inferencing models run on purpose-built silicon: NVIDIA Jetson for computer vision, Google Coral for TensorFlow Lite workloads, or custom FPGAs for latency-critical applications. In the near edge — a local gateway or server — aggregation, preprocessing, and more complex analysis runs on x86 hardware. In the cloud — the centralised layer — model training, large-scale analytics, and long-term data storage live where unlimited compute is available.
5G is the infrastructure catalyst for edge computing. Its combination of high bandwidth and ultra-low latency allows more workloads to be executed at the network edge rather than requiring on-premise servers. As 5G coverage expands in Indian cities and industrial zones, edge computing applications that required dedicated local hardware will increasingly be deployable on telco edge infrastructure.
The industries with the most immediate edge computing opportunity in India are manufacturing (quality control, predictive maintenance), retail (autonomous checkout, inventory tracking), and healthcare (real-time patient monitoring, in-clinic diagnostics).