Network traffic has traditionally moved in one dominant direction: from networks toward users. The internet is currently experiencing a shift from consumption to production.

Streaming video, social media, cloud applications and online advertising have made downlink capacity the central concern for mobile operators. Networks have largely been designed for people who consume digital content through smartphones, tablets and computers.

Artificial intelligence could challenge that model. As AI moves beyond text-based applications and into smart glasses, connected vehicles, industrial systems and robotics, devices will increasingly generate data rather than simply receive it. Cameras, microphones and sensors will capture the physical environment and send that information upstream for analysis. 

The response from the network may be relatively small: an instruction, recommendation or inference result.

This emerging pattern could change the direction, timing and performance requirements of mobile communications. Instead of downloading a video or webpage, a user may rely on an AI system that continuously receives live audio, video and sensor information. The network would become less like a content delivery system and more like a distributed nervous system.

At Mobile Future Forward, T-Mobile President of Technology John Saw said AI has not yet produced a significant change in mobile network usage. Most of the impact remains concentrated in transport networks and around large data centers. Still, Saw said he expects the effect to grow as AI workloads move closer to the edge.

Uplink Demand Could Accelerate

According to the Ericsson Mobility Report, uplink traffic has been measured at approximately 8% of total traffic across a sample of networks. Conventional video traffic has been even more unbalanced, with roughly 97% moving downstream and only 3% traveling upstream.

Changes are expected when AI systems gain eyes, ears and other sensors. Ericsson

Research indicates that cloud-based AI and augmented-reality smart glasses could require between 1 Mbps and 10 Mbps of uplink capacity. Some applications could produce a downlink-to-uplink ratio of approximately 1:8 as cameras, microphones and sensors transmit information to cloud-based AI systems.

Ericsson’s projections indicate that uplink traffic could reach three times 2025 levels by 2031.

Overall mobile data traffic, by comparison, is forecast to grow by approximately 2.2 times during the same period.

Operators will not merely need more capacity. They may need a materially different balance between downlink and uplink resources.

Physical AI Raises the Stakes

The network challenge becomes more complex when AI is applied to machines operating in the physical world.

 Robots and humanoid systems face limits involving battery life, onboard computing power and model size. Offloading some processing to the network could allow machines to use more advanced models while reducing local power consumption.

A robot cannot always wait for a distant data center to process a request. A delayed response could reduce efficiency or create a safety concern.

Network Traffic Volume Rises

AT&T Business Product chief Shawn Hakl has described an architecture in which users may rely on six to 12 AI models, with those models distributed across devices, enterprise sites, edge facilities and multiple clouds. 

Using Haki’s model, data may move through several systems before an answer returns to the device.

The Counterpoint Research analysis of AI and mobile traffic similarly anticipates increased upstream movement of images, voice, video and contextual data. Smart glasses, industrial equipment and machine-to-machine agents could all contribute to this growth.

Additional context is available through the USA Herald Jackie Allen author search, including coverage of telecommunications, artificial intelligence, legal issues and digital infrastructure.

Planning Before the Network Traffic Surge

The immediate threat and anticipated complications should not be overstated. 

The GSMA Mobile Economy reports estimate that direct generative AI accounts for only a small share of mobile traffic today. Operators could also overbuild capacity if they invest too aggressively before AI applications reach mass adoption.

Nevertheless, network architecture must be planned years before demand arrives. Spectrum allocation, radio configuration, transport networks, edge computing and 5G-Advanced capabilities all require long investment cycles.

Mobile networks were optimized for a world in which people downloaded the digital environment. AI could create a different world, one in which billions of intelligent endpoints continuously describe the physical environment upstream. 

If that transition develops at scale, the most important mobile traffic question may no longer be how much content people want to download. It may be whether networks are prepared for machines that are constantly sending the world to AI.

See additional articles in the USA Herald by Jackie Allen . This includes coverage of topics like  telecommunications, artificial intelligence, legal issues and digital infrastructure.