Sir Airffic GoThis links to smartphones and game consoles, so think of everyday examples.
Why compute on the aircraft?
Avoiding obstacles in flight needs a very short gap between seeing and deciding. There is no time to send data to the cloud and wait.
Drones must also fly where radio does not reach or is jammed. So critical decisions run on the aircraft's own chip. This is edge AI.
A strict budget of battery and weight
The heavier a drone is, the shorter it flies. The more power the computer draws, the shorter it flies too.
Chips that do the same work with less electricity are worth more to drones. The key number is not peak speed but computation per watt.
Sir Airffic GoWhen you compare chips, do not just pick the fastest. Look for speed that is kind to the battery.
Chipmakers sell the whole toolbox
NVIDIA offers Jetson compact AI computers together with robotics software and simulators. Once developers learn those tools, switching chips becomes harder.
Qualcomm brings power-efficient chips and communications honed in smartphones to drone development platforms.
Chip companies want to own the development platform, not just sell parts — much like the Windows-and-Intel relationship of the PC era.
What changes with foundation models
Large models that understand images and language are still too heavy for most drone chips.
Expect a split: shrink models to run on board, keep urgent decisions on the aircraft, and send slower decisions to the ground or cloud.
Deciding where each computation runs is exactly the seam between the onboard brain and the airspace OS — a big opportunity for operations platforms.
Frequently asked questions
What is edge AI?
Running AI on the device itself — a drone or camera — rather than on cloud servers, which reduces the effect of network delays and dropouts.
Which AI chips do drones use?
Examples include NVIDIA's Jetson series, Qualcomm's Flight/RB platforms and Ambarella's camera chips. Companies such as ModalAI build drone-specific compute modules on top of them.