Will AI Provide a Boost in Wireless Traffic as we Move Towards 6G?

The telecommunications industry is contemplating a problem it hasn't faced before: a lack of growth. As reported by IDTechEx "6G Market 2026-2036: Technology, Trends, Forecasts, Players", YoY growth rates in global wireless traffic have been on a downward trend since 2020 (albeit still exhibiting growth). As the industry continues to lay the groundwork for the CAPEX-intensive shift to 6G, could various manifestations of AI offer a pathway to renewed growth?

 

 

Ever since the first wireless services provided analogue calls in the 1980s, the number of raw bits transmitted through global wireless networks has been rising steadily. As more users were added to the network and the nature of data transmitted grew more complex. Analogue voice calls became texts, which became full HD streaming, and this inexorable rise in traffic has fuelled the telecoms industries growth and ushered in successive 'generations' of wireless transmission. While Q1 of 2025 saw a record 171 EB of total uplink and downlink traffic, the trailing YoY growth rate had fallen to under 20%. For contrast in 2019 the same growth rate peaked at over 90% (and was regularly above 50%). What is behind this slowdown in growth?

 

The evolving nature of network traffic can clearly be seen in the statistics. In 2012 video streaming took up 32% of the overall traffic, and by 2024 this had ballooned to 74% making it by far and away the most significant application for wireless networks. Video is a far more data-intensive form of media than either text or audio, and the relentless rise of clip-based social media as well as conventional streaming have driven much of this growth. However, could this growth be hitting a fundamental limit? IDTechEx analysis estimates that a user that streams video for 4 hours a day, every day, at 1080p would average around 150 GB/month. This would be 4 hours a daystreaming video and would not account for any other Wi-Fi based content consumption.

 

In short, the amount of traffic that can be derived from video streaming faces 3 constraints:

 

With handheld and tablet devices unlikely to drastically increase in size, resolutions are limited. There are only 24 hours per day, and people must sleep, work, travel, and as such have a fundamentally limited amount of time in which to stream content. While developing countries are continually adding new users to networks, most developed countries have already saturated the number of people who are connected.

 

This 150 GB ceiling is impressive and significantly above the current global average, but it is an edge-case to illustrate that even with intensive streaming at high resolutions there is a limit to how much growth can be derived from video traffic alone. In other words, to find the renewed growth that the industry seeks, a new untapped source of traffic must be found.

 

 

Telecoms, like almost every other industry, is searching for ways to capitalize on the AI boom and find tangible and monetizable benefits. While there are many angles' telecoms is pushing for AI integration (including AI network automation and optimization), one of the more interesting arguments is that AI will provide a new burst of traffic to networks and allow growth just at the time growth rates have been slowing down.

 

On the surface, this seems to make sense. The explosion in computational power and data associated with hyperscalers could be expected to translate into an enormous amount of new data being transmitted between cloud and edge users, which would require increased telecoms infrastructure to support this. However, nearly all use-cases of AI are less intensive than streaming. For instance, a typical ChatGPT query consists of a short text prompt and text response. While a significant amount of computation may be required, the amount actually transmitted across the wireless network to the user is fairly small. The bulk of data transmission and processing occurs in giant datacentres - which do face huge bandwidth bottlenecks (leading to innovations in optical transceivers and co-packaged optics) - the data sent to users is plain text.

 

With some of the most intensive use-cases of today, generative AI video, the amount sent to devices is at most high-definition video and audio, which is not a new type of data for wireless networks. A shift to on-device processing would further mitigate the burden on networks, further nullifying the argument that AI will drive wireless traffic up.

 

Whilst incumbent use-cases of AI are unlikely to drive mobile traffic back towards increasing growth, there are potential use-cases that could change the dynamics. Enormous quantities of training data, mostly in the form of data-intensive video, could be sent from self-driving cars back to datacentres to train and inference new models for autonomous driving. However, the huge spike in data transfer requirements by AI are most likely to be felt in data centres and core networks, rather than by a strain on bandwidth at the network edge. For more information on the unfolding shape of 6G, see "6G Market 2026-2036: Technology, Trends, Forecasts, Players".

 

For more information on this report, including downloadable sample pages, please visit www.IDTechEx.com/6G, or for the full portfolio of related research available from IDTechEx, see www.IDTechEx.com.