Artificial intelligence returns home: why home AI is challenging the cloud

Home AI sometimes goes hand in hand with the smart home. A well-known example here is Home Assistant Green (by Nabu Casa). It is a PC-based platform that manages devices and automations whilst keeping data within the home. This is in contrast to traditional smart speakers such as Alexa.

Users can connect lights, sensors, cameras, thermostats and smart sockets to a system that continues to function even if the internet connection is lost. Home Assistant is also introducing voice and AI features that can be run locally.

The device retails at around 179 euros on the European market and is probably the easiest way to get started in the world of home automation without having to assemble hardware or configure servers. Some advanced features require additional components, but the basic experience is designed for a much wider audience than the traditional ‘maker’ community.

The same concept of taking back control of one’s data is evident in the field of personal storage. BeeStation, developed by Synology, offers an alternative to the traditional cloud: photographs, documents and backups remain physically in the home, whilst retaining the ease of access typical of online services.

The device is available in 4- and 8-terabyte versions and starts at around 414 euros on the European market. The idea is to offer an experience similar to Google Drive or Dropbox without entrusting one’s files entirely to external servers. Some of the automatic image cataloguing features utilise artificial intelligence techniques.

The final piece of the jigsaw is computers designed specifically to run AI models locally. This is where Framework Desktop comes in: the modular, repairable PC that has caught the attention of local AI enthusiasts.

Based on AMD Ryzen AI Max processors, the system can support up to 128 gigabytes of unified memory, a particularly attractive feature for running large language models. Prices start at around 1,429 euros for the base configuration, but rise rapidly with additional memory, storage and components.

One advantage of these solutions is that they allow you to replace at least some of your subscriptions with a one-off investment in hardware. But there is also the satisfaction of being able to control your own data and services without having to go through third parties (namely the American big tech companies).

The limitations are clear. On-premises AI requires greater technical expertise, maintenance and initial investment. Larger models still require cloud infrastructure, and many services remain easier to use when they are fully managed by third parties. Yet the appeal of home (or office) AI is strong. Especially for those who come from a different kind of internet – one that is more diverse and under the control of each and every one of us.