Edge Computing

Moving computing power to where the data is created — and why it is reshaping industries from manufacturing to autonomous vehicles.

The global edge computing market is projected to reach $110 billion by 2028, growing at 37% CAGR, as organisations recognise that sending all data to centralised cloud data centres is too slow, too expensive, and increasingly untenable. By 2026, an estimated 75% of enterprise-generated data will be created and processed outside traditional data centres — at the "edge" of the network. For business leaders, edge computing represents a fundamental shift in IT architecture that enables new revenue streams, reduces costs, and unlocks capabilities impossible with cloud-only approaches.

What It Is

Edge computing refers to processing data close to where it is generated — on factory floors, in retail stores, on mobile phone masts, in vehicles, or on IoT devices — rather than transmitting it to a centralised cloud data centre for processing. The "edge" is wherever the data originates and where immediate action is needed.

Consider a self-driving car. It generates approximately 40 terabytes of data per hour from its cameras, lidar, and radar sensors. Transmitting all that data to a cloud data centre for processing is impossible — not only would the bandwidth costs be astronomical, but the round-trip latency of even 100 milliseconds could be the difference between avoiding a collision and causing one. The car must process sensor data locally, in real time, and make driving decisions within milliseconds. This is edge computing in its purest form.

Edge computing does not replace the cloud — it complements it. The typical architecture is a three-tier model: edge devices handle real-time processing and immediate decision-making, regional edge nodes (sometimes called "fog" nodes) aggregate and pre-process data from multiple edge devices, and the central cloud handles long-term storage, model training, and deep analytics. Data that needs millisecond responses stays at the edge; data that can wait seconds, minutes, or hours flows to the cloud.

Why It Matters for Business

Edge computing matters because it solves three problems that cloud-only architectures cannot: latency, bandwidth cost, and data sovereignty. Latency-sensitive applications — autonomous vehicles, industrial robotics, real-time video analytics, augmented reality — require response times under 10 milliseconds. Even the fastest cloud data centres cannot guarantee this when the data must travel hundreds of kilometres over network infrastructure. Edge processing eliminates this network latency entirely.

Bandwidth costs are another powerful driver. A single factory with 1,000 IoT sensors generating 10 GB of data per day would incur $150,000–300,000 in annual cloud data transfer costs alone. Processing data locally at the edge and sending only aggregated insights to the cloud can reduce data transfer volumes by 90%, cutting connectivity costs proportionally. General Electric reported saving over $100 million annually by processing industrial IoT data at the edge rather than transmitting raw sensor data to the cloud.

Data sovereignty regulations — including GDPR in Europe, data localisation laws in China, India, and Russia, and sector-specific regulations in financial services and healthcare — increasingly require that certain data be processed and stored within national or even organisational boundaries. Edge computing provides a natural architecture for compliance, keeping sensitive data on-premises whilst still enabling cloud-based analytics on anonymised or aggregated data.

Enterprise Use Cases

Walmart deploys edge computing infrastructure across its 4,700+ US stores, processing data from point-of-sale systems, inventory sensors, and security cameras locally. This enables real-time inventory management — automatically triggering restocking orders when shelves run low — and AI-powered loss prevention that identifies suspicious behaviour in real time. Walmart reports that edge-based analytics have reduced out-of-stock incidents by 30% and improved loss prevention accuracy by 50% compared to centralised cloud processing.

In telecommunications, Vodafone has deployed Multi-Access Edge Computing (MEC) infrastructure across its European network, enabling enterprise customers to run applications directly on Vodafone's mobile base stations. A manufacturing customer uses this capability to run real-time quality inspection AI directly at the factory floor, processing images from production line cameras with 5-millisecond latency — fast enough to flag defective products and trigger automated removal before they progress further down the line.

Boston Dynamics uses edge computing on its Spot robotic inspection dogs deployed in industrial environments. Spot processes lidar, thermal imaging, and acoustic sensor data locally to navigate complex environments and detect equipment anomalies (gas leaks, overheating machinery) in real time, sending only exception reports and summary data back to the cloud. This edge-based approach enables autonomous inspection rounds lasting 90 minutes without network connectivity.

Vendor Landscape

The edge computing market is fragmented across infrastructure, platforms, and applications. AWS offers AWS Outposts (starting at approximately $7,000 per month for a single rack) and AWS Snow Family for edge deployments in remote locations. Azure provides Azure Stack Edge (from $3,000 per device) and Azure IoT Edge for lightweight edge workloads. Google Distributed Cloud offers edge variants for both connected and disconnected deployments. specialised edge vendors include Fastly and Cloudflare (compute at CDN edge locations), Rigado and Litmus (industrial edge platforms), and VMware Tanzu (edge orchestration). Hardware vendors Dell (PowerEdge XR series), HPE (Edgeline), and Lenovo supply ruggedised edge servers designed for factory floor and outdoor deployment, typically priced 20–40% above equivalent data centre hardware due to hardened specifications. For AI at the edge, NVIDIA Jetson modules (starting at $149 for entry-level devices up to $1,999 for the Jetson AGX Orin) are the dominant platform, with over 1 million units deployed in commercial applications.

Strategic Considerations

Invest in edge computing when your use cases demand real-time response (under 100 milliseconds), when bandwidth costs for transmitting raw data are prohibitive, or when regulatory requirements mandate local data processing. Do not pursue edge computing for workloads that tolerate latency well — standard cloud architectures are simpler, cheaper, and more manageable for these scenarios.

The primary challenge of edge computing is management complexity. Deploying and maintaining software across hundreds or thousands of geographically distributed edge locations requires robust orchestration, remote management, and security patching capabilities. Invest in edge management platforms early, and establish clear policies for edge device lifecycle management, security updates, and data governance. Ensure you have a strategy for handling intermittent connectivity — edge devices must operate reliably even when network connections are unstable or temporarily unavailable. Budget for ongoing operational costs, which can be 2–3 times higher per unit of compute at the edge compared to centralised data centres.