Nokia’s AI Applications Study: “Physical AI” may require RAN redesign to support high‑volume, low‑latency uplink traffic

Nokia said an analysis of Artificial Intelligence (AI) applications in mobile networks pointed to changes in traffic patterns that could affect Radio Access Network (RAN) design. The company’s review focused on how AI-generated traffic behaves now and what network behavior physical AI would require as it becomes part of mobile traffic. That framing matters for operators planning capacity, service performance, and network architecture.

In Nokia’s account, most mobile networks show AI-generated traffic at an early stage, with application maturity and adoption by consumers and enterprises only starting a broader AI super cycle. Nokia said it examined more than 50 AI applications and found three outcomes: higher uplink traffic, overall data growth, and increasing sensitivity to delay in conversational services such as chat and voice. The company also cited industry movement toward “AI-RAN” or “6G-native” structures that embed AI into the network.

Nokia described mobile networks as supporting heterogeneous traffic, with operators typically meeting capacity increases through infrastructure expansion and overprovisioning under best-effort delivery. It said voice shifted from circuit-switched patterns to packet-switched IP traffic, and that LTE-M and Narrowband Internet of Things (IoT) (NB-IoT) developed to address Massive Machine-Type Communication (mMTC). For physical AI, Nokia contrasted buffering-based delivery used for video streaming with applications that require strict time-to-live constraints for high-definition video frames and sensor data, shifting the focus toward consistent low latency and guaranteed scheduling.

Nokia’s analysis argued that meeting physical AI requirements could mean abandoning best-effort models in favor of reserved capacity or specialized AI-RAN functionalities, and it connected that to uplift video and latency targets. The company said delivering uplink video with sub-20 ms end-to-end latency can require provisioning three to four times the average uplink capacity, making overprovisioning costly for high-throughput video streams. It cited a multi-layer approach spanning network architecture, traffic management, and service monetization, and it referenced Quality of Service (QoS) and network slicing, plus semantic communication for reducing transmitted data. “Physical AI introduces the possibility that large-volume uplink video with strict latency requirements. It will become a meaningful part of mobile traffic, creating both a design challenge and a monetization opportunity,” says Harish Viswanathan, Head of the Radio Systems Research Group at Nokia.