The Death of Data Growth vs. Math, Technology and AI

When I saw last week’s news about Wireless Data Growth in the US (up 35% between 2024 and 2023) my initial reaction was, “sure, that tracks.” More 5G devices. More coverage. More FWA. Stable pricing. Put it all together and the results aren’t surprising. Then came my follow-on reaction. 

“How long until we hear that the sky is still falling and mobile data growth is still headed towards zero at some point in the future?” 

It’s a classic strategy of pundits and proto analysts who are invested in their worldview. Forecasts and predictions are rarely wrong. Originally estimated timelines, however, are subject to change. In the long run, we all die; we can’t be sure it’s true, but it’s hard to argue with. 

Similarly, in the long run, we cannot be certain of where wireless network traffic growth will go and, to be fair, slowing growth rates can be extended into a simplistic forecast which leads to flat or declining traffic. As last week turned into this week, however, I was reminded why this isn’t a worldview I find convincing. 

• Smaller Changes vs. Bigger Numbers. The first thing most of us learn about saving for retirement is the power of compound growth. Apply that to network traffic and you arrive at a stark reality; even if traffic growth rates taper, the amount of traffic is growing from a bigger and bigger base. Framed differently, 50% of 10 is much smaller than 20% of 100. The latter scenario implies slower growth, but much more traffic for networks to carry.
• Tomorrow’s Applications. Not long after the US wireless traffic news, there was an announcement of 5G Broadcast getting demoed at IBC this year. When it popped up as an in-demand 5G-Advanced use case, we were all surprised here at GSMA Intelligence – broadcast/multicast over 4G, and 3G have both gotten attention, then been dismissed given the potential drag on network capacity. I’ve heard similar pre-emptive concerns about ICAS. While not the same as capacity demand it all represents added network capacity requirements.
• Today’s Applications (plus). A common reply to the suggestion that wireless data traffic will continue growing long into the future goes something like this. “Video is the dominant form of traffic. People cannot watch more than one video at a time. Ergo, there’s a clear limit to growth.” Fair enough…maybe. While lots of people are watching lots of video, most of us aren’t watching that much. GSMA Intelligence consumer survey analysis, for example, indicates that in the UK, France and Germany, fewer than one-in-four adults watch video on a daily basis. This is reflected in the traffic data, where 10% of heavy mobile users account for 60-70% of total traffic. It’s possible/probable that the aforementioned industry pundits are over-represented in this group, but looking outside this bubble shows there’s plenty of room to grow video’s impact on the network, especially if we see usage driven by quality upscaling, more uplink video and conversational apps which drive video in both directions. Which leads us to…
• The AI Question Mark. We know that the mainstreaming of AI will impact network traffic. We just don’t know the direction. Lots of multi-modal queries? Traffic goes up. Proliferation of IoT devices (sensors) to generate data for AI models? More traffic. Agents talking to agents talking to agents? Traffic goes up, at least marginally. AI content scaling and conversational AI built into the apps we’re already using? Yep, you guessed right, traffic goes up. AI content optimization and video compression? Traffic tumbles, potentially by a lot. On-device AI processing AI workloads, well, on the device? AI-driven traffic increases are tempered. 

This last bit is obviously a wildcard. Based on the number of “traffic goes up” vs. “traffic goes down” scenarios I’ve listed, you can guess where my head is at. Regardless, it’s only one part of the equation and, in the here and now, the data shows robust growth on big numbers with no shortage of ideas for how to keep networks busy. Whether or not you identify as a chicken, the sky isn’t falling.