Login

Cart

Your cart is empty

Physical AI and AI Factories: Building the Intelligence Behind Smart Cities

Updated on: November 04, 2026 · By Melike Gülsunay Çalışkan

Physical AI brings intelligence into systems that sense and operate in the real world, while AI factories provide the large-scale computing foundation used to develop, train, simulate, and optimize AI models. In smart cities, edge AI connects these layers by bringing deployed intelligence closer to cameras, sensors, machines, and transportation infrastructure.

 

Smart cities are becoming complex physical environments in which roads, buildings, transportation systems, cameras, sensors, and machines continuously generate information. The challenge is no longer only connecting these assets. Cities also need AI infrastructure that can transform physical-world data into useful intelligence and support AI systems after they move from development into deployment.
Title

What Is Physical AI?

Physical AI refers to artificial intelligence that can perceive, understand, and interact with physical environments using information from cameras, sensors, vehicles, machines, and other connected systems. In a smart city, this can include interpreting traffic movement, understanding conditions around transportation infrastructure, supporting autonomous machines, or analyzing sensor information from facilities.

 

What Is an AI Factory?

An AI factory is computing infrastructure built for developing, training, optimizing, and running AI models at scale. It combines accelerated computing, networking, storage, software, and data infrastructure so teams can work with demanding AI workloads. For smart city projects, it can support the central part of the lifecycle before models move into distributed deployment.
Title

How Do AI Factories, Edge AI, and Physical AI Work Together?

These technologies solve different parts of the same problem. AI factories provide scale for model development and optimization. Edge AI provides proximity to physical data sources. Physical AI describes the intelligence operating within or alongside real-world systems.

  • Central development: models are trained, tested, simulated, and optimized using large-scale AI infrastructure.
  • Edge deployment: selected models run near cameras, sensors, machines, or transportation systems.
  • Physical operation: AI interprets real-world conditions and produces outputs for a defined application.
Title

Where Can Physical AI Support Smart City Infrastructure?

Physical AI can support intelligent transportation infrastructure, smart buildings, autonomous machines, environmental sensing, and infrastructure inspection. The computing requirements can differ significantly between these use cases, which is why smart city AI infrastructure is usually distributed rather than built around one computing platform.

 

These applications can span transportation networks, public spaces, buildings, utilities, and other connected environments. In each case, AI systems need to process information from the physical environment and turn it into insights or actions that can support city operations.

Title

How Do Digital Twins Fit Into the Physical AI Lifecycle?

Digital twins can represent roads, buildings, transportation networks, machines, or other environments in software, allowing teams to simulate scenarios before physical deployment. NVIDIA Omniverse is one example of a platform used to build and simulate digital environments for Physical AI.

Why Edge Computing Matters for Physical AI

Physical AI applications often depend on information generated directly in the environment. Processing selected workloads at the edge can reduce the need to continuously transfer large volumes of raw data to centralized infrastructure and can make local information available closer to the application. Edge computing does not replace the AI factory; it extends AI infrastructure into the physical environment.
Title

How Forecr Connects Physical AI to the Edge

Forecr develops NVIDIA® Jetson™-based carrier boards and industrial computers that provide the local computing foundation for edge AI deployments. In Physical AI applications, this hardware layer connects AI models with cameras, sensors, machines, and other systems operating in the physical environment. Different deployments can require different combinations of processing performance, connectivity, interfaces, storage, power, and thermal design. 

 

Forecr platforms provide developers and system integrators with hardware options for bringing NVIDIA Jetson™ computing into these real-world environments.

For smart city applications, this can support distributed AI architectures where selected workloads operate closer to the physical systems generating data.

Title

What Should Developers Consider When Deploying Physical AI?

Moving an AI model from development into a physical environment introduces requirements that may not appear during model training or simulation. The deployment platform needs to support both the AI workload and the systems surrounding it.

Key considerations can include:

  • Processing requirements: The selected platform should provide sufficient compute and memory for the models used in the application.
  • Sensor and camera connectivity: Physical AI systems may need to receive information from multiple cameras, sensors, or machines.
  • Networking: Ethernet and other communication interfaces can be important when edge systems exchange information with wider infrastructure.
  • Power and thermal conditions: Deployment environments may have different power limits, airflow conditions, or temperature requirements.
  • Physical integration: Form factor and available interfaces can influence how the computing platform is integrated into a vehicle, cabinet, machine, or other system.

These requirements help determine what type of edge computing platform is appropriate for a Physical AI deployment.

Title

From Connected Infrastructure to Physical Intelligence

Smart cities have spent years connecting cameras, sensors, buildings, transportation systems, and other infrastructure. Physical AI represents another step in this evolution by bringing AI models closer to the environments they are designed to understand.

 

AI factories can provide the computing scale required to develop, train, simulate, and optimize these models. Edge computing extends this infrastructure into the physical world, where AI can process information from local cameras, sensors, machines, and connected systems.

 

Together, these layers create a path from connected infrastructure to physical intelligence, linking AI development with real-world deployment across smart city environments.

Need help?

Frequently Asked Questions

Melike Gülsunay Çalışkan

Melike Gülsunay Çalışkan | Marketing Specialist

With nearly five years of experience in marketing, brand communication, and e-commerce analytics, I focus on building clear, consistent brand narratives for technology companies...