One of the consequences of artificial intelligence (AI) systems becoming capable of reasoning, planning and executing tasks autonomously is that mobile traffic patterns are changing noticeably, with uplink being of growing importance.
A study from InterDigital has shown how the emergence of agentic AI will redefine the demands placed on devices, networks and cloud infrastructure. Among the findings of the The distributed network shift enabling AI on device report, conducted by ABI Research for InterDigital, was that the rapid adoption of agentic systems – which is expected to increase across enterprise and consumer markets over the next three years – was increasing uplink traffic from AI devices, changing the way modern networks operate. The result will be a reimagining of network design.
The study noted that modern mobile networks have historically been optimised for downlink throughput and video delivery. However, unlike traditional mobile applications that primarily consume data via downlink, agentic AI systems continuously generate and exchange contextual information to enable real-time reasoning and decision-making. Therefore, as AI devices generate increasing volumes of upstream data, networks risk becoming overloaded, leading to higher latency and costs.
The study found four main devices driving uplink traffic: smart glasses, wearables, smartphones, IoT sensors and devices. Smart glasses continuously capture video, images and environmental context, sending data upstream for real-time AI inference and assistance. ABI Research predicts 70 million smart glasses shipments by 2030, with cellular-enabled devices representing more than 12% of shipments.
By contrast, wearables – including next-generation tech that collects voice, biometric and contextual signals – support persistent agentic AI interactions. Smartphones increasingly transmit multimodal inputs such as voice, photos, video and sensor data to cloud and edge AI systems. In their operation, IoT sensors and devices continuously stream operational or environmental data to AI models for analysis, automation and decision-making.
The study also found uplink pressures are already visible in video-heavy applications such as livestreaming and real-time video collaboration, where many users uploading simultaneously can create localised mobile cell congestion. It added that unlike these temporary spikes, agentic AI systems will generate continuous upstream data exchanges from connected devices, potentially creating sustained pressure on uplink capacity.
The report suggested that to meet AI demands of modern devices, the industry must transition toward distributed intelligence architectures, where AI workloads are orchestrated across on-device processors, and cloud platforms based on their complexity. It said that embedding intelligence deeper into network infrastructure will ensure AI-enabled applications can operate efficiently without compromising on performance.
The study observed that as the entire mobile ecosystem continues to innovate and integrate the latest AI technology at pace, ensuring a coherent and complementary direction of travel is essential to enabling future AI applications and their associated experiences.
This is particularly seen as the case for 6G networks, which will be designed to make smartphones better at Mobile Broadband (MBB) access by improving network speeds, reducing latency and refining the battery life of devices.
However, InterDigital cautioned this is just the foundation on which additional services will be built. Integrating AI in the network will allow smartphones to offload demanding applications to the edge of the network – as well as into centralised locations – to ensure optimal resource utilisation, enabling a distributed intelligence fabric.
“Agentic AI introduces a new set of requirements for both networks and devices,” said Larbi Belkhit, and Paul Schell, senior analysts at ABI Research and co-authors of the report. “Supporting autonomous AI systems will demand far more distributed computing architectures and significantly more intelligent networks. Operators will need to manage increasingly symmetrical traffic patterns while enabling real-time AI workloads across device, edge and cloud.”
“Agentic AI marks the next phase in the evolution of intelligent connectivity,” said InterDigital chief technology officer Rajesh Pankaj. “Intelligence must be distributed across devices, networks and the cloud, and delivering these AI-enhanced services efficiently will require a new computing architecture that balances performance, latency and energy efficiency.”
