Cities and infrastructure operators across Europe are testing locally run AI agents and digital twins for building maintenance, transport, safety and public services. The projects use Nvidia models and computing, but range from exploratory pilots to deployed systems, so their maturity and evidence vary.
In France’s Côte-d’Or department, agents combine sensor, infrastructure and historical records across more than 500 public buildings. Nvidia says the system reduced document searches from weeks to minutes and cut operating costs by 10 percent, with a 30 percent target at larger scale. In the Milan region, a local AI installation is being deployed for translation, citizen chatbots and municipal procurement assistance.
Other projects are earlier. Liverpool, Jersey and Belfast are exploring systems for mobility, planning and emergency response. The Port of Antwerp-Bruges is testing synthetic training data for oil-spill warnings, while Gothenburg is developing a digital twin that previews entrances, seating and crowd conditions for accessibility planning.
The common technical approach combines video search, language models and simulated environments, with hardware operated locally so public bodies retain control of their data and models. The announcement comes from Nvidia and highlights customers using its stack, rather than an independent comparison. Results from one building portfolio should not be assumed for other cities, and several named deployments are still trials rather than daily public services.