Light Reading
Michelle Donegan
January 22, 2026
A burning question for mobile operators is how AI will impact their networks. They worry that an onslaught of AI generated traffic from smart glasses and smartphones could overload networks not designed for uplink capacity. Results from a new study by Signals Research Group show the impact is not as great as some expect and might put minds at ease.
Signals Research Group set out to measure the effect of AI and augmented reality (AR) applications on cellular data and signaling traffic in a 5G network. It tested Meta Ray-Ban Display glasses and two different Samsung Galaxy smartphones, and all the AI applications they support as well as some AR apps.
The field tests were conducted on T-Mobile’s 5G standalone network using test equipment from Accuver.
On the key uplink capacity question, the result is that uplink traffic from image- and video-based generative AI is indeed “much higher” than text, but the impact is “very modest.”
“There’s a lot of concern [about uplink]. I won’t say it’s not valid but it’s overblown … Yes, there’s more bandwidth being used in the uplink, but on a relative basis compared to other things, it’s trivial,” said Michael Thelander, CEO and founder of Signals Research Group.
Further, he noted that improvements to AI and AR user experience are possible without needing to make changes in cellular networks, as the apps become more sophisticated.
That doesn’t mean operators should take investment and focus away from 5G uplink capacity. Thelander said there is still a need to improve network uplink in 5G and with new features coming in 6G. “But we’re not killing the networks with what we have today with AI and AR,” he said.
For a comparison on how networks are holding up for other heavy data apps, he pointed out that video calling on cellular networks has been around since 2012 and that there are 150 million iPhones in the US that each can do FaceTime calls, but no one is worried about it because the overall time spent on video calls is “modest.”
“If you look at the progression for what can happen with AI and AR over the next several years, I don’t see there being a huge impact, especially compared to the FaceTime example – [it] generates so much bandwidth, and nobody cares about that … I’m not saying we ignore [AI and AR]. I’m just saying, take a deep breath and relax, it will work itself out,” said Thelander.
Signaling surprise
An unexpected finding was that Meta Ray-Ban smart glasses and Meta AI app generated a whopping amount of signaling traffic. Signaling was not in the initial scope of the tests, but it was added after Thelander noticed some odd traffic behavior.
Even when the glasses were off and the app was closed, the smartphone remained attached to the network, keeping a radio resource control (RRC) connection, and sending signaling control messages. The amount was “very high,” said Thelander.
“There is a significant amount of transactions where the phone is connecting to the network for no good reason, and then, depending on the phone model that I tested with, it was staying connected for a considerable amount of time,” he said.
The result was baffling and Thelander refrained from drawing broader conclusions about whether this is a problem or not, or if it is a glimpse of a possible “signaling storm 2.0” scenario.
“A single device generates a heck of a lot of signaling traffic. Back in the days of 3G, it would have crushed the network. 5G is much more efficient when it comes to signaling, so I don’t know where that all plays out,” he said, noting that the signaling impact would be greater if a user was not connected to a 5G SA network.
Back around 2010, so-called signaling storms hit AT&T’s and others’ 3G networks associated with the iPhone and “chatty” apps.
AI drain on smartphone battery
Signals Research also tested AI apps that run natively on smartphones, without a connection to the cellular network. Not surprisingly, these apps are power hungry and drain the battery.
Thelander said the battery impact is “significant,” but it’s not likely to be a problem unless the apps are used for long periods at a time.
“The intelligence you can get on your phone is impressive. There’s a significant counter impact in terms of what it does to your battery, but it’s such a short duration … that it’s probably not something you’re going to be overly concerned about,” he said.
Neglected uplink now in focus
With each advance in network technology or addition of new spectrum, mobile operators tend to prioritize boosting download speeds to support all the apps and services people want to use. Less attention, and capacity, is given to uploads.
Ookla lead industry analyst Mike Dano has analyzed how operators around the world are allocating uplink capacity in his latest research, given the potential for AI to disrupt traditional traffic patterns with a steady stream of uploads.
The study found that while upload speeds have risen between 2021 and 2025, operators have not been increasing the portion of capacity allocated to uplink connections in the same period.
Looking at 17 operators, China’s three mobile operators dedicated the biggest percentage of capacity to the uplink connection and US operators allocated the least, according to Ookla.
However, Dano noted that while AI could put new demands on networks, “there are plenty of unknowns in this emerging space.” Notably, it is not certain to what extent a cellular connection will be needed if much AI usage is processed on devices by apps running natively on smartphones.