In our MWC25 Barcelona predictions webinar (check it out to see what we got wrong), we suggested AI would be the show’s dominant theme. We were right, of course.
I cannot say it was a particularly risky guess. Riskier, however, was the prediction we would begin to see the AI narrative evolve from one focused primarily on AI for networks to include a more meaningful networks for AI story. It was somewhere between a forecast and an aspiration, but we got it right nonetheless, a good thing given that we decided to run an Executive Roundtable at MWC on the topic. More importantly, between the news from the show and our own industry engagements, we came away with key insights into how operators are thinking about building networks to support AI workloads.
If you aren’t familiar with the AI for Networks versus Networks for AI dynamic, it is a straightforward conceptual framework. On the one hand, AI can be used to support network operations; think proactive maintenance, RAN optimisation, or genAI-driven troubleshooting. On the other hand, telecoms networks will be critical for supporting AI workloads and AI innovation going forward and we will need to increasingly build or evolve those networks with AI in mind.
To date, AI for Networks has dominated much of the industry’s focus. This is logical to the extent operators have been employing AI in their network operations for years. But it potentially ignores what could be a key in executing on the broader AI opportunity, the networks and connectivity which will enable AI applications, along with an opportunity for operators to monetise their networks.
It was good, then, to see Networks for AI get plenty of attention at MWC.
From transport network and data architecture upgrades, to AI-focused edge networking and 5G-Advanced launches, the notion of network upgrades being made with an eye towards AI use cases was front and centre at MWC. In addition, announcements from operators pointed to the sheer breadth of ways in which networks will need to support AI’s growth, extending all the way to user devices and home gateways at the far edge of the network.
Of course, tradeshow news is one thing, but the true value of an event like MWC is the opportunity for engagement with decision makers and innovators. This is why GSMA Intelligence took the opportunity to hold our Networks for AI Executive Roundtable (Intelligence Hour: Building Networks for Mobile AI – What Will It Take?) with leading infrastructure suppliers, RAN innovators, operators and analysts at MWC. As we work to understand how, exactly, networks will need to evolve to support AI, the frank and open discussion was invaluable and wide-ranging.
A trio of key takeaways, however, stood out.
- Data traffic implications. While mobile data traffic growth rates have slowed as smartphone and mobile broadband adoption nears saturation in many developed markets and no new application categories have emerged to drive traffic in the way mobile video once did, AI could represent a new disruption. There is a very real concern AI traffic, particularly content-rich, multi-modal queries and responses, could generate new traffic demands and stress networks, especially on the uplink. As processing of AI workloads gets increasingly distributed, it is not clear which parts of the network will be most stressed, but a need for more RAN capacity is assumed, making the case for 5G-Advanced upgrades.
- Physical AI and the link to IoT. It is well understood AI is fuelled by data. The amount and quality of data fed into a model drives the quality of the results. It is equally well understood AI’s impact will touch the digital as well as physical domains. We already see this in terms of innovation around robotics or the development of digital twins covering everything from manufacturing plants to entire communities. Generating the data to support physical AI aspirations will require economical and widespread sensor deployment at scale. Again, technologies like 5G-Advanced (including Ambient IoT) and RedCap were cited as possible solutions.
- Monetisation Imperative. 5G network monetisation has been a perennial theme of MWC since the onset of the 5G era. It was the driving force between a focus on 5G in support of enterprise verticals. It is responsible for new B2C business models and tariffs based on experience differentiation. It is not surprising, then, that our Roundtable participants, including operators which serve more than a billion connections, cited monetisation as key to any network investments they make in support of AI use cases. While it should be obvious operators would look to earn a solid return from any new network investments, it is a reminder Networks for AI represent a revenue growth opportunity as well as an AI necessity.
– Peter Jarich – head, GSMA Intelligence
The editorial views expressed in this article are solely those of the author and will not necessarily reflect the views of the GSMA, its Members or Associate Members.