What Is Edge Computing?

Why your data does not always need to go to the cloud — and why sometimes it cannot afford the round trip.

↓

How It Works

Edge Computing: 6 Key Concepts

From local processing to cloud sync — how the edge works.

🔌

1. Processing at the Source

Tap to flip

Instead of sending all data to a central cloud server, processing happens near the device that generates it — on a local server, gateway, or even on the device itself. The “edge” is anywhere between the device and the cloud.

⚡

2. Low Latency (Speed)

Tap to flip

Cloud round trips take 30–100ms. Edge processing can be 1–5ms. For self-driving cars, industrial robots, or real-time video analysis, that difference is the difference between working and crashing. ^[1]

🔒

3. Privacy & Bandwidth

Tap to flip

Processing data locally keeps sensitive data on-site — important for healthcare, finance, and surveillance. It also reduces bandwidth costs since only processed results (not raw data) are sent to the cloud. A factory sensor generating 1TB/day cannot upload everything.

🏗️

4. Edge Devices & Gateways

Tap to flip

An edge device is the thing generating data (sensor, camera, phone). An edge gateway aggregates data from multiple devices and does local processing. Example: a router that runs AI models to detect network threats before they reach the cloud.

⚙️

5. Edge AI (Inference at the Edge)

Tap to flip

AI models can now run directly on edge devices (Apple Neural Engine, Google TPU, NVIDIA Jetson). Your iPhone processes Face ID and Siri voice recognition locally — no cloud call needed. Edge AI is the fastest growing segment of computing in 2025–2026. ^[2]

📶

6. Cloud + Edge: Together

Tap to flip

Most architectures are hybrid: edge for real-time processing, cloud for heavy training and storage. Your smart thermostat processes locally (edge) but syncs usage patterns to the cloud for long-term optimisation. Cloud trains the AI; edge runs it.

Compare

Cloud vs Edge vs Fog

Three computing models — each suited for different needs.

☁️

Cloud Computing

Tap to flip

Data is sent to centralised data centres for processing. High latency (30–100ms), high bandwidth costs, but unlimited compute power and storage. Best for training AI models, batch processing, and global applications. AWS, Azure, GCP.

🌍

Edge Computing

Tap to flip

Data is processed near the source — on-device or on a local gateway. Ultra-low latency (1–5ms), lower bandwidth, better privacy. Best for real-time applications: self-driving cars, factory automation, AR/VR, video surveillance.

🌪️

Fog Computing

Tap to flip

An intermediate layer between edge and cloud. Fog nodes (local servers) aggregate data from multiple edge devices, process it, and send condensed results to the cloud. Think of it as a local data centre for a factory or building. ^[3]

🏗️

Example: Smart Factory

Tap to flip

Sensors on a production line generate terabytes per day. Edge gateway runs AI to detect defects in real-time (5ms). Only defect alerts + summaries are sent to the cloud. Bandwidth cost drops 99%. Downtime reduces from hours to seconds.

🚗

Example: Self-Driving Car

Tap to flip

A self-driving car generates ~5TB of sensor data per hour. It cannot send that to the cloud and wait for a response. The car processes everything locally (edge). Only anonymised driving statistics are uploaded later for model training.

📱

Example: Your Smartphone

Tap to flip

Face ID, voice typing, and live translation all run on-device (edge). Siri and Google Assistant process simple requests locally. Complex queries go to the cloud. This is edge computing that billions of people use daily without knowing it.

Quiz

Edge Computing Knowledge Check

6 statements. TRUE or FALSE?

About where edge computing processes data:

Edge computing processes data on centralised cloud servers.

About the main benefit:

Edge computing reduces latency compared to sending all data to the cloud.

About the cloud+edge relationship:

Edge computing will eventually replace cloud computing entirely.

About AI at the edge:

AI models can run directly on edge devices like smartphones without a cloud connection.

About fog computing:

Fog computing is the same as edge computing.

About data volume:

A self-driving car generates enough data per hour that sending it all to the cloud is impractical.

Scenarios

When Do You Need Edge Computing?

From convenience to necessity — slide to see when edge is the right choice.

050100

Slide to assess

Key Takeaways

Edge Computing: What to Remember

Tap to check off each concept.

Edge = processing near the source

Instead of sending all data to the cloud, edge computing processes it on local devices or gateways.

Latency is the main driver

When every millisecond counts (self-driving cars, robotics), cloud round trips are too slow. Edge is the only option.

Edge AI is growing fast

Modern phones, cameras, and sensors can run AI models locally. Face ID, voice recognition, object detection — all at the edge.

Cloud + edge, not cloud vs edge

The best architecture uses both: edge for real-time decisions, cloud for training and storage.

Fog is a middle layer

Fog computing aggregates multiple edge nodes on a local network before sending to the cloud.

Privacy improves at the edge

Sensitive data (medical, financial, surveillance) stays local. Only anonymised results go to the cloud.

Bandwidth savings can be huge

Edge processing can reduce cloud data uploads by 90–99%, dramatically cutting bandwidth costs.