The push to run large AI models directly on devices — rather than in distant data centers — is quickly shifting from a nice-to-have to a hard requirement, according to a report from 36Kr.
The outlet frames the trend around a single pointed question: edge-side large models have become a "rigid demand," but are the chips actually ready to deliver on it?
"Edge-side" refers to AI that runs locally on the hardware in your hand or on the factory floor — phones, cars, cameras, and other gadgets — instead of sending every request over the internet to be processed remotely. Running a large model this way promises faster responses, better privacy, and the ability to work without a constant connection. The catch is that squeezing a demanding model onto a small, power-constrained device puts intense pressure on the underlying chips.
That pressure is the heart of 36Kr's story. If devices are expected to handle serious AI on their own, the processors inside them need to advance in step — a challenge the report suggests is not yet fully resolved.
The source item does not detail specific chipmakers, product names, or performance benchmarks, so the broader picture is a question being posed rather than a verdict being delivered: demand for on-device AI is racing ahead, and the silicon has to catch up.
Why it matters: whether chip capabilities can match the surging appetite for on-device AI will shape how quickly — and how well — the next wave of smart products actually works in people's hands.