Physical artificial intelligence, which runs on next-generation telecom networks to move things in the real world, is expanding rapidly from controlling individual devices to managing city infrastructure and industrial sites. Major global carriers are racing to lead the smart city and smart factory markets with base stations that carry AI computing power and ultra-low-latency networks.
T-Mobile, one of the three largest U.S. carriers, is piloting a network-based physical AI platform in San Jose, California, and using it to manage the city's traffic network, according to foreign media reports and the telecom industry on the 14th. At the core of the platform is an "operating agent" that runs directly on base stations installed in the field. AI agents inside the base stations analyze data gathered by roadside cameras and sensors in real time to immediately detect unexpected traffic accidents. When an accident is spotted, the base station adjusts nearby traffic signals or identifies alternative routes on its own and relays the information to vehicles in the area, without routing the request through a central control center. Telecom infrastructure effectively becomes a giant control tower that monitors citywide traffic flows in real time and intervenes in the physical world.
In Japan, a manufacturing powerhouse, major carriers are also leading physical AI demonstrations in robotics and manufacturing with next-generation networks. SoftBank is combining base stations with multi-access edge computing, or MEC, to advance control technology for autonomous robots operating in factories and offices. The approach improves the performance of multiple robots already deployed on site simply by updating AI software installed on the network, without replacing individual robot hardware. The company is also accelerating demonstrations with Yaskawa Electric, a leading Japanese industrial robot maker, to embed physical AI in manufacturing sites that require precise physical tasks.
The NTT Group has also moved to secure an early lead in the manufacturing AI ecosystem with IOWN, a next-generation optical communications technology. Using an ultra-high-speed network that transmits large volumes of data as light, sharply cutting power consumption and latency, the group is running demonstrations that tie computing resources at remote data centers hundreds of kilometers away to industrial robots on factory floors as if they were a single neural network. The strategy is to achieve precise remote control and process automation through the network, without installing expensive semiconductors directly in on-site equipment.
In Europe, telecom equipment maker Nokia has built an industrial fifth-generation, or 5G, test environment together with the University of Oviedo in Spain and automation specialist Festo. Robots recognized products on a conveyor belt with cameras, used AI-based image analysis to decide which items to pick up, and then moved them to a packaging line. The setup uses a private 5G network and edge computing to support tasks that link mobile robots with production equipment.
The information technology industry expects such demonstrations to fundamentally reshape carriers' business-to-business and business-to-government revenue models, because they could allow carriers to emerge as core platform operators running vast physical spaces such as smart factories and smart cities in real time. "For physical AI to settle into daily life and industrial sites, it needs networks that process data from countless devices and city infrastructure without a moment's delay and turn it into action," an industry official said. "Competition among global carriers to upgrade infrastructure will intensify as it moves from robots to the city level."