The business shifts focus towards agentic AI and sovereign infrastructure, based on the GSMA
MWC 2026 was full of AI messaging, which is not any shock. This time round nonetheless, issues appear to be shifting a bit — from idea to one thing a bit extra concrete. In a dialog with RCR Wi-fi Principal Analyst Sean Kinney, GSMA Intelligence Head Peter Jarich painted an image of an business that’s lastly crossed over from the “artwork of the potential” to the “artwork of the sensible.” The best way Jarich sees it, the AI dialog at MWC has crystallized into three distinct lanes — agentic AI, sovereign AI, and AI-RAN. The hazard, as he notes, is that the business is pouring its vitality into the flashiest factor — AI-RAN — whereas neglecting the stuff that’s already producing actual returns, like agentic AI, AI utilized to core community operations, and sovereign AI deployments.
Agentic AI strikes from idea to deployment
Solely a yr in the past, the telecom business’s tackle AI brokers was, generously, aspirational. Jarich recalled how the dialog went again then. “You’ll hear some tech execs describe AI brokers like drunk interns and it’s form of you already know, does it present up? Sure or no, perhaps. If it does present up, do you actually know what it’s doing? Does it know what it’s doing? Can it inform you what it’s doing? Would you belief it with something?”
Quick ahead to MWC 2026 and the image appears to be like fully completely different, based on the GSMA. The agentic AI summit was packed, and operators weren’t pitching hypotheticals anymore. They had been sharing what they’d realized from precise deployments. Kinney, who was on the summit, put it merely: “this was all very a lot the artwork of the sensible, not the artwork of the potential.” Jarich backed that up: “We’ve had a yr of learnings and so over the past yr folks sort of figured put it to make use of.”
In keeping with Jarich, although, the fascinating half isn’t essentially the tech itself — it’s about the whole lot surrounding the tech. He drew a pointy line between two basically completely different postures towards AI adoption. “Take an current enterprise course of and the way do you layer AI on prime of it to make it higher?” versus “What can AI do and what would I do beginning with AI and constructing on prime of AI versus layering AI on prime of what I’m doing.”
That first strategy, which mainly includes bolting AI onto current workflows, will get you effectivity enhancements. There’s nothing mistaken with that, but it surely’s incremental. The second strategy, nonetheless, includes designing totally new processes and companies, round AI. “There’s going to be a world of distinction whenever you hear right here’s what I’m doing and the way do I make it higher versus how do I construct one thing new on prime of AI, how do I rethink, create new companies,” Jarich mentioned.
He additionally acknowledged a extra uncomfortable dynamic enjoying out at some corporations, the place dramatic workforce reductions are getting used primarily as a forcing perform to set off that deeper organizational rethink. “Typically whenever you hear these massive issues the place you’ve received folks going I’m going to chop half my workers, they’re doing that as a result of they wish to drive that perform. As soon as I lower half my workers, everybody’s going to need to go oh I now must rethink the whole lot we do.” He was fast so as to add the caveat: “I’m not saying that’s the most effective concept, I undoubtedly don’t suggest it to my boss. Don’t eliminate half my workers.” However the broader level stands — extracting actual worth from agentic AI calls for a real rethinking of how operations and companies are constructed, not simply headcount optimization.
Sovereign AI as a geopolitical alternative
Whereas agentic AI is basically an organizational readiness problem, sovereign AI is rooted in geopolitical actuality. Jarich flagged it as one in all MWC 2026’s standout themes, describing a rising push by governments worldwide to take possession of their AI and cloud infrastructure stacks. “How do governments wish to deploy not simply AI however cloud companies, compute companies. Plenty of them have needed to determine how they put them to make use of for their very own functions, perhaps develop language fashions folks won’t, develop their very own communities, ship companies for governments,” he defined. “We additionally need to admit given the place we’re with geopolitical conditions, they wish to have some extra management over their very own destinies on that facet of issues.”
The examples are already stacking up. France is a very hanging case. “A number of weeks in the past France determined that going ahead 2027 and past authorities companies gained’t be capable of use the coms options, like Groups, Zoom, they’ll want to make use of an in-country developed resolution,” Jarich famous. Then, weeks later, France directed {that a} government-built shopper well being information platform be migrated off Microsoft Azure. “I take a look at these as simply examples of how governments are fascinated with what sovereign means,” he mentioned.
For telecom operators, this opens issues up a bit. Telcos carry a stage of belief, they know native regulatory environments inside and outside, and so they already function inside nationwide boundaries. However Kinney pushed on whether or not this chance has an expiration date — whether or not hyperscalers may ultimately muscle into these markets and type out the regulatory panorama themselves. Jarich conceded the window gained’t be open indefinitely however challenged the belief that hyperscalers will blanket each market: “It’s unclear if hyperscalers can go into each market, proper? So I believe some markets undoubtedly, some workloads sure, however not essentially all of them and that can rely market by market.”
Geography is the necessary variable right here. Sovereign AI isn’t strictly a Western European phenomenon. “The worth of those sovereign options aren’t simply going to be a worth to a Germany or a UK or a South Korea but in addition to a Pakistan or a Kyrgyzstan or a Somalia or markets the place the hyperscalers could not see the chance or the worth and even when they do, they’re most likely not going to go and develop language fashions for all these international locations,” Jarich mentioned.
The AI-RAN definition downside
AI-RAN was nonetheless the large matter at MWC 2026. The anticipated gamers dropped a wave of bulletins. However peel again the press releases and issues get muddier quick. The business, Jarich argued, can’t even decide on a standard definition of what AI-RAN really is — and that confusion is an actual drag on coherent progress.
“For an Ericsson, who did a bunch of bulletins earlier than the present, for them a few of AI-RAN is likely to be I’m going to place AI in my radios to enhance radio efficiency. And you could possibly inform like from a number of the analysts that I used to be speaking to, they’re sort of like wait, after I assume AI-RAN that’s not what NVIDIA was speaking about, proper?” Jarich mentioned. “However it’s nonetheless utilizing AI in your RAN, it’s simply not utilizing AI processors to run baseband. It’s good as a result of it exhibits the alternative ways you may apply AI. It’s dangerous as a result of I believe it doesn’t make it simple for telcos to completely wrap their head round it.”
However the greater situation, in Jarich’s view, is what all of the AI-RAN hype is drowning out. “What there aren’t questions on is how I can apply AI within the core, proper? As a result of there’s a number of core community workloads that you just already can apply and individuals are making use of AI to that may generate revenues, enhance companies, lower OpEx,” he mentioned. “I believe as a lot as AI-RAN is sort of horny and it’s all the eye, sort of wish to see a bit bit extra consideration to the AI within the core as a result of it’s right here now, and we will do that whereas we’re determining the remaining.”
Sustainability and 6G sensing
Outdoors the three most important AI lanes, Jarich known as out two areas he felt deserved way more airtime than MWC 2026 gave them.
On the sustainability entrance, the dialog has narrowed in ways in which clearly frustrate him. Inexperienced community discussions confirmed up on the present, however they had been wrapped up when it comes to OpEx discount not when it comes to sustainability. Jarich pointed to the GSMA Intelligence’s Inexperienced Community Index, which has grown from six taking part operators at launch to 24 this yr, as an try and push the body wider. “We’d been taking a look at like vitality effectivity for telcos for some time, proper? And saying that’s one a part of the story however you even have to have a look at energy like information heart energy effectivity use of renewables. In addition to similar to precise community efficiency as a result of I can run a brilliant energy-efficient community which doesn’t ship a lot when it comes to efficiency.” Operator curiosity is clearly there; the business’s public framing simply hasn’t caught up but.
On 6G, Jarich was genuinely stunned that Built-in Sensing and Communications (ISAC) didn’t command extra of the highlight. He famous that at a 3GPP 6G assembly in Korea shortly after final yr’s MWC, ISAC and precision positioning ranked among the many prime precedence areas for each operators and distributors. At this yr’s present, GSMA Intelligence hosted a roundtable on the subject and “the curiosity was simply off the charts.” What makes ISAC stand other than a number of 6G conversations is that it’s not only a 5G use case ready for extra bandwidth — it “feels undoubtedly new,” as Jarich put it.
