Linker Vision's AI Aims to Power Autonomous Smart Cities

SAN JOSE, CA – March 16, 2026 – A future where city infrastructure can perceive, reason, and act autonomously is taking shape at NVIDIA's GTC 2026 conference, with Linker Vision showcasing a platform that aims to turn this vision into reality. The AI company is highlighting its Video Reasoning AI platform, an ambitious system designed to serve as the intelligent backbone for smart cities and spaces, with a pilot project already underway to test its capabilities in San Jose.

Already deployed in major Asian cities including Kaohsiung, Taiwan, Linker Vision's technology moves beyond simple video analytics. It promises to create “city operation agents”—AI systems that provide contextual understanding of live events, analyze potential impacts, and support complex decision-making. In San Jose, the company is working with partners Voxelmaps and Inchor to test digital twin traffic simulations, exploring how these agents can optimize urban flow and improve incident response in a real-world environment.

The Blueprint for an Autonomous City

At the heart of Linker Vision's platform is a deep integration with NVIDIA's full stack of AI technologies, transforming vast streams of video data into what the company calls “agentic intelligence.” This is the core of Physical and Reasoning AI: systems that don't just identify objects but understand context, causality, and complex scenarios.

To achieve this, the platform leverages the NVIDIA Metropolis Blueprint for video search and summarization (VSS). This framework allows for the analysis of massive-scale video feeds, enabling operators to summarize hours of footage in minutes or search for specific events in seconds. Layered on top are NVIDIA's Cosmos open world foundation models (WFMs), which provide the crucial reasoning capability. These models interpret complex visual scenes, allowing the system to understand not just what is happening, but why it might be happening and what its implications are.

This isn't merely theoretical. In Kaohsiung, one of Taiwan's largest cities, Linker Vision has integrated 30,000 diverse camera streams into a city-scale 3D digital twin. The system is trained to understand over 300 scenarios across ten urban domains, from traffic management to disaster response. City officials report that leveraging the NVIDIA VSS blueprint has slashed development efforts by 85% and, more critically, improved incident response times by up to 80% by breaking down information silos between departments.

From Digital Twins to Real-World Action

Training an AI to manage the near-infinite variables of a bustling city presents a monumental data challenge. It's impossible to capture real-world footage of every conceivable event, especially rare but critical “long-tail scenarios” like a multi-vehicle pile-up at a specific intersection during a flash flood. Linker Vision addresses this through a digital twin-enabled training pipeline aligned with the new NVIDIA Physical AI Data Factory Blueprint, also announced at GTC.

This reference architecture is a game-changer for developing AI for the physical world. It automates the generation, curation, and validation of training data at scale. Using Cosmos foundation models like Cosmos Transfer, the blueprint can take limited real-world data and generate vast, diverse sets of high-fidelity synthetic data. This allows Linker Vision to train its models on a far wider range of events than would be possible otherwise, ensuring the AI is robust and reliable when deployed.

To manage this data-intensive process, Linker Vision is testing its toolchain on Microsoft Azure, integrating the NVIDIA blueprint with Azure's scalable cloud services. This collaboration provides the massive computational power needed to run the data factory, continuously generating model-ready training datasets from raw data and accelerating the entire development lifecycle. The San Jose pilot, which aims for a fivefold improvement in incident response, is a direct beneficiary of this advanced training methodology, enabling its AI agents to learn and simulate scenarios before they ever happen on the city's streets.

The AI Grid: Telecoms Join the Smart City Revolution

For an AI to act in real-time, processing must happen instantaneously. Sending petabytes of video data to a distant cloud data center introduces unacceptable latency for critical applications like traffic control or emergency response. The solution is to bring the intelligence to the source—a concept being realized through “AI grids.”

Linker Vision is at the forefront of this shift, forging key partnerships with major telecom operators to build geographically distributed AI infrastructure at the network edge. With AT&T, the system runs on edge sites that act as AI grid nodes, bringing video reasoning capabilities closer to the cameras and sensors themselves for faster, more reliable operations. This aligns with AT&T's broader strategy of building a “Connected AI” platform for the industrial edge.

An even more forward-looking collaboration is being piloted with T-Mobile. Here, Linker Vision's platform is being deployed on a grid of AI-RAN (Radio Access Network) ready infrastructure powered by NVIDIA's latest RTX 6000 PRO Blackwell Server Edition. This initiative, highlighted by NVIDIA CEO Jensen Huang at GTC, aims to transform standard wireless cell sites into powerful AI computing platforms. “In the future, [the radio tower is] going to be an AI infrastructure platform — a robotics radio tower that can reason about traffic,” Huang stated, underscoring the transformative potential of embedding AI directly into the fabric of the 5G network.

A New Competitive Edge in a Crowded Market

Linker Vision operates in a vibrant and increasingly crowded market. NVIDIA's own Inception program supports over 800 AI startups focused on smart city applications, all vying to provide the next generation of urban management tools. However, Linker Vision has carved out a significant strategic advantage.

Its unique selling proposition lies not in a single feature, but in its comprehensive, end-to-end platform that seamlessly integrates the most advanced components of the AI world: NVIDIA's full software and hardware stack, a sophisticated digital twin and synthetic data pipeline, and crucial partnerships with telecom giants to deliver intelligence at the edge. This holistic approach moves beyond piecemeal solutions to offer a unified, scalable operating system for physical spaces.

This strategic positioning has been validated by NVIDIA itself, which participated in Linker Vision's $35 million Series A funding round in January 2026. This investment is more than just capital; it's a powerful endorsement of Linker Vision's technology and its central role in NVIDIA's ecosystem for building intelligent digital twins worldwide.

“As digital twins operate everywhere, reasoning AI can scale with them—unlocking new opportunities for smart cities and critical infrastructure,” said Paul Shieh, Founder and CEO of Linker Vision. By connecting simulation, training, and deployment into a continuous AI lifecycle, the company is building a tangible framework for bringing Physical and Reasoning AI into real-world operations.