NVIDIA Expands City AI With Shared Infrastructure and Digital Twins

NVIDIA says it and its partners are expanding physical and agentic AI services for cities worldwide, with projects spanning infrastructure planning, transportation and public services.[1] The shared-computing approach is intended to let city departments, neighboring municipalities and local developers extend services without each acquiring and managing its own resources.[1] NVIDIA says the required equipment depends on the workload: an RTX workstation or a small GPU system may be enough.[1] It also says the approach lets local governments retain control over how sensitive data is processed and used.[1] Together, those points make workload sizing and data governance central considerations for organizations evaluating a deployment.

In Spain, 1MillionBot combines open-source ALIA models trained at Barcelona Supercomputing Center with NVIDIA Nemotron and other models in agents; NVIDIA reports more than 150 deployments across the country.[2] The services include municipal procedures, appointment requests, urban-planning incidents and service inquiries.[1] LiveX offers navigation, personalized recommendations and conversational assistance in airports, hospitals, retail stores and public venues, and says it is deployed at more than 100 customer sites.[3] Its named AI City locations include San José Mineta International Airport, El Camino Hospital, SAP Center, PUMA stores, and Marriott and Hilton hotels.[3] NVIDIA says LiveX’s conversational AI Hosts can be tailored to local languages and service needs.[1]

NVIDIA says LiveX AI City uses Cosmos world models, Nemotron open models, Omniverse and Metropolis libraries on NVIDIA RTX PRO GPUs.[1] A related system is designed to turn existing camera streams into operational information for mobility, safety and event response across several city functions.[2] ITMAX bundles infrastructure and agentic AI applications, including traffic management, into long-term managed services for cities.[1] ITMAX says its SamurAI V2 platform is rolling out across more than 17,000 cameras, with the full fleet planned for 2027.[4] The company reports average peak-hour travel times 26% shorter on 10 Kuala Lumpur corridors where adaptive signals are live.[4] It also reports improved alert verification and faster incident searches.[4]

Linker Vision unveiled Linker AI Nexa on November 3, 2026, connecting real-time city data with government standard operating procedures.[2] NVIDIA says Nexa uses Metropolis VSS and Cosmos to interpret real-world conditions, Omniverse libraries and OpenUSD for digital twins, and Nemotron reasoning models for natural-language scenario exploration.[2] Linker Vision’s environmental, social and governance agent can simulate 15-year canopy, shade and thermal-comfort outcomes in 10 minutes; the work can support multiple design iterations in an hour, compared with approximately two weeks per iteration.[2] Separately, Honeywell’s Forge Plant Cognition uses Nemotron-powered agents to interpret operational data at Dangote’s Lekki refinery, with an Omniverse-based digital twin planned for simulation and training.[2]