MORAIDICTIONNAIRE IA
Infrastructure

Edge AI (Edge AI)

Running AI directly on the device itself, without relying on the cloud.

Picture a miniature brain tucked directly inside your watch, your car, or your security camera: that is the idea behind Edge AI. Instead of shipping data to a remote server for analysis, the artificial intelligence model runs locally, as close as possible to the sensor producing the data — the so-called edge of the network.

Why bring AI closer to the data

Cloud computing requires a costly network round-trip. Edge AI removes that detour, delivering three key benefits:

Total latency breaks down simply:

$$ T_{\text{total}} = T_{\text{network}} + T_{\text{compute}} $$

Running locally, $T_{\text{network}} \approx 0$, yielding instant responsiveness.

The cost of the constraint

Squeezing a model into a chip drawing a few milliwatts forces trade-offs. Engineers rely on quantization (going from 32 to 8 bits, or fewer), pruning, and distillation to shrink the network.

Criterion Cloud (datacenter) Edge AI
Latency High Very low
Privacy Data sent away Data stays local
Compute power Near-unlimited Limited
Energy High Low

Edge AI is intelligence moving toward the data, rather than the other way around.

Explore the full AI dictionary →