- Local on-device AI offers privacy, zero latency, and no API costs, but is largely impractical for most consumer devices.
- Consumer hardware faces RAM constraints, with 8-16GB common, leaving only 4-8GB for AI models after OS and apps.
- A capable 7B parameter model requires ~5GB, but agentic tasks also need large KV cache, quickly exceeding 8GB RAM.
- Even small agentic tasks need at least 32K token context, requiring ~16GB RAM just for AI on devices.
- Rising RAM prices and competition with datacenter HBM worsen memory availability for consumer devices.
- On-device inference speed drops drastically with context size, e.g., from 100 to <10 tok/s at 16K context on a desktop GPU.
- Battery and heat issues make sustained on-device agentic AI use impractical on phones.
- Cloud inference avoids these limits but would require enormous datacenter capacity for billions of devices.
- Realistic agentic AI on consumer devices remains years away without major hardware or architecture breakthroughs.