The tech industry has often been defined by massive shifts in behavior that upend conventional wisdom about consumer preferences and how our communications infrastructure supports them.

When the first iPhone was launched in 2007, networks were unprepared for the massive download demands it sparked. We may be approaching a similar turning point, where AI transitions from being a feature within our devices to becoming embedded in the hardware itself. Network operators must ensure they are prepared for this shift and avoid being caught off guard. Smartphones fundamentally changed how people used mobile devices, shifting from phone calls and texting with some occasional web browsing to constant connectivity, app ecosystems, and rich media.

Now, the explosion in AI adoption is setting the stage for a new era of devices - ones that will not follow the traditional mold of screen-based phones or computers. With AI built in from the start, tomorrow's devices will be omnipresent, always-on, and optimized for new connectivity challenges as they collect and process ambient data from their surroundings.

This shift has the potential to upend the traditional balance between downloading and uploading data. Unlike apps and streaming services, which rely on large data downloads, AI-centric devices will increasingly feed information back into the network. This fundamental change will place unprecedented strain on uplink capacity, requiring networks to evolve.

Historically, networks have been optimized for downstream traffic. Think about how your cell phone plan is marketed – carriers usually emphasize blazing-fast download speeds for gaming, streaming, and simultaneous device activity, with uplink speeds often treated as an afterthought.

The good news is that we can learn from the past. This time around, we have a better idea of what to expect and just how quickly paradigm shifts can happen. With a wave of AI-centric devices expected to hit the market next year, it’s time for networks and operators to prepare for new demands.

AI-centric use changes user traffic

In the last 20 years, we’ve seen a gradual buildup of data traffic with a new wave of devices driving user behavior. Emerging technologies like AR/VR, while not yet mainstream, are gaining traction, used for everything from immersive gaming to remote collaboration and creating an insatiable demand for bandwidth. Now, AI is rapidly accelerating change in both the habits of consumers and the data flow inside mobile networks.

Down the line, the AI experience, currently focused on browsers and apps, could evolve into wearable consumer AI agents, and enterprise devices could collect operational or environmental data to feed AI models continuously. And it’s not just consumers driving this change, enterprise use cases like AI-enabled sensor networks in warehouses or ports are a factor.

These use cases create a two-way data stream, continuously uploading movement, audio, and environmental inputs to sync with Edge compute or cloud-based platforms - and they depend on high-bandwidth, low-latency connections.

As consumers grow more comfortable with these experiences, prices for devices and tools drop, and new hardware rolls out, this behavior will only increase - and users will expect seamless, responsive interactions. Networks have to be prepared and can’t get stuck playing catch-up.

Data demand will put a new emphasis on the uplink

As mentioned earlier, networks have traditionally been optimized for download traffic. However, AI-centric devices - particularly screenless, voice-first, or wearable ones - will reverse that priority. Use cases like constant data collection, voice inputs, ambient sensing, and video streaming will put an unprecedented strain on uplink capacity.

Currently, there are only a handful of situations where uplink demand exceeds downlink, like a stadium where an athlete scores a goal, and thousands of people rush to post videos of it online. AI devices will create a pervasive, “always-on” version of this effect, as both enterprise and consumer behavior evolve.

To meet this demand, network resource allocation must shift toward uplink, moving away from the 80/20 mix networks today tend to be planned around for media-consumption.

There’s also an element of spectrum allocation that is critical here. Even today, mobile operators are looking for more spectrum to be released to meet the demands of an exploding ecosystem of connected devices. AI devices will need access to additional spectrum bands for capacity and performance, adding further complexity to network design and operations.

Networks will evolve from communications to include sensing

Increasing uplink demand won’t be the only shift. As opposed to focusing on content delivery, the Edge AI experience will rely on constant collection of ambient data, including sound, motion, and proximity. That real-time data will then be processed locally and at the Edge, integrating networks into a cyber-physical sensing convergence.

For example, the network itself can add integrated sensing which can be used to detect and track drones that are at an altitude and provide assistance to the UAV traffic management system for navigation and collision avoidance.

To enable these use cases, networks will need to adapt to support integrated sensing and communications (ISAC), which brings sensing and spatial locations of passive (not connected) objects into the mobile communication network.  

Preparing network infrastructure

We’re at the edge of a new computing shift. AI-centric devices may change the way mobile networks are built just as radically as the smartphone did - and operators should prepare to stay ahead of the curve as user behavior shifts, so they’re not left at a competitive disadvantage – or unable to keep up with demand.

First and foremost, spectrum allocation must shift toward uplink needs, and mobile use cases that will rely on ISAC need to be considered in network planning. On top of these shifts, there’s a short runway for preparation, as new devices are predicted to make an impact as soon as 2027 as they roll out for widespread consumer adoption.

Changes will also be called for as CSPs look deeper into the network infrastructure away from the Radio. A shift to cloud-native architecture is necessary to support the scalability, flexibility and reliability needed to support such advanced technology as AI -driven devices. A high-performance, expandable, cloud-native 5G Core network with good Edge capacity, as well as upgraded transport and routing, will be essential for CSPs wishing to provide a competitive, attractive, economically successful service for these devices.

CSPs also should consider the telecom IT aspect, the supporting applications and services that ensure the network runs smoothly. Upgrading and updating this IT backbone to a cloud-native architecture that can handle real-time or near-real-time processing and orchestration across a number of key areas like charging, analytics, service creation means CSPs are able to not only handle the quantity of data the network needs to handle thanks to the AI-driven devices, but can also be easy to buy from and get paid as a service provider and will have a technical set-up that can stay ahead of business as it rapidly changes.

The groundwork must begin now. As the cyber and physical worlds converge, devices are no longer simply operating as part of a communication network but truly becoming something more. Ambient sensing, reversed traffic patterns, and Edge AI is the new mandate – and this time, we can see what’s coming and prepare.