AI dominated the conversation at Mobile World Congress 2026, but the more interesting story was about who controls the infrastructure it runs on.
I came back from Mobile World Congress 2026 in Barcelona with a lot to say. I always do.
For context: MWC is enormous in a way that is difficult to convey to someone who hasn't been. 110,000 attendees. 2,900 exhibitors. The exhibition floor stretches roughly 1.5 kilometres. If you took Times Square, moved it indoors, and then kept going for another 20 minutes of walking, you'd start to get the idea. It is loud, relentless, genuinely exhilarating, and also deeply useful if you know what you're looking for.
I attend as someone who works closely with the ISP and broadband community. That's the lens I bring to three themes that I think deserve attention: AI's operational role in telecom networks, the open standards movement gaining momentum, and the geopolitical conversation around digital sovereignty that, frankly, should be making more people uncomfortable.
AI in telecom is not the AI you're thinking of
AI was the dominant theme at MWC 2026. The terms "AI native" and "AI Ops" were everywhere. But I want to make an important distinction, because the AI being discussed in Barcelona is different from the AI conversation happening everywhere else.
When most people talk about AI right now, they're talking about large language models: the tools that write, summarise, and generate.
The AI being embedded in telecom networks is about designing networks so that AI systems continuously monitor, optimise, and control network behaviour. Practically speaking, that means:
- Predicting network outages before they happen
- Self-healing networks that identify and fix problems before a single customer notices
- Dynamic spectrum management
- Anomaly detection in traffic patterns
It means your ISP's network infrastructure getting smarter and more autonomous in ways that are almost invisible to end users (which is exactly the point).
There was also significant discussion about moving AI computation to the network edge rather than routing everything through centralised hyperscale cloud infrastructure. Edge data centres, base station compute platforms, and regional telecom clouds are becoming the locus of AI processing for anything latency-sensitive: industrial automation, robotics, autonomous systems, augmented and virtual reality. For ISPs, this has direct implications for where compute capacity gets deployed and how network architecture needs to evolve.
That said, I want to temper this with some honesty. There is a lot of space between what companies say they're doing with AI and what they are doing. MWC is a marketing event as much as a technical one, and everyone there is putting their best foot forward. I think we are genuinely seeing important developments, but I'd expect it to take a few more years before we can assess which of these announcements translate into real, operational change at scale.
Open Telco AI and Open RAN: two sides of the same coin
One of the most interesting announcements at MWC 2026 was the Open Telco AI Initiative from the GSMA (Global System for Mobile Communications Association), the industry body that sponsors MWC. The goal is to develop AI models specifically trained on telecom data and standards. This matters because the general-purpose large language models available today genuinely struggle with telecom-specific requirements. Network configurations, radio optimisation, telecom protocols, operational troubleshooting: these are domains where sector-specific models should meaningfully outperform general ones.
This is particularly interesting to us because companies like AT&T, Orange, Vodafone, and Telefonica (all InkBridge Networks customers) have signed on to this initiative. AT&T is contributing to open telecom AI models while AMD is providing compute infrastructure. The shared goal is to avoid dependence on the hyperscaler AI platforms and build AI capabilities that are specifically optimised for network operations. As a company deeply committed to open standards and open source, I find this development genuinely exciting.
Running alongside this, and closely connected, is Open RAN - an open standards initiative for Radio Access Networks (RAN), the wireless access layer of a mobile network. This is the technology that connects phones, laptops, and IoT devices to the telecom operator via radio signals. Traditionally, RAN systems are closed and vertically integrated within a single vendor's ecosystem. Open RAN is trying to break that dependency.
For more on how InkBridge approaches open standards in network authentication, our work on RADIUS 1.1 gives a good sense of how we think about open protocol development.
The familiar tension is present here too: everyone in the industry wants global interoperability, except the large vendors who would prefer to own the entire stack. Open Telco AI and Open RAN are parallel movements toward an open model for AI orchestration running on open infrastructure - and together, they represent a meaningful challenge to the closed-ecosystem playbook. I think it's exciting. I also think we should wait five years before declaring victory.
Digital sovereignty, and a conversation that surprised me
I also want to highlight something that I found genuinely alarming at MWC 2026, and I don't use that word lightly.
Digital and data sovereignty was a major theme among European operators and vendors. Given the current geopolitical environment, and the well-documented shift in US technology policy, this is understandable.
European organisations are taking a serious look at their infrastructure dependencies on US-based hyperscalers and AI platforms.
I had multiple conversations with European vendors who are actively discussing moving away from US AI hyperscalers and considering Chinese ones as a replacement.
I had to make sure I heard that correctly.
Some of the vendors being considered as alternatives are explicitly identified as supply chain security risks in Canada, the US, and significant parts of Europe. What that designation means in practice is that any entity considered critical infrastructure - telecommunications, utilities, anything essential to running a country - is legally prohibited from using equipment or software from those vendors. Not discouraged. Prohibited by law. These are vendors that have been determined to be actively collecting data from their customers and reporting it to their government.
So the proposed solution to concerns about data sovereignty is to replace one set of infrastructure dependencies with another set of infrastructure dependencies - from vendors that are already the subject of supply chain security bans in the very countries proposing to use them.
I understand that the geopolitical environment is genuinely difficult and that organisations are trying to navigate real and legitimate risks. But swapping one sovereignty concern for one that has already been acted on legally highlights the fact that the entire global IT infrastructure is currently completely dependent on two countries. And neither of them feel like safe bets right now.
If this is a conversation happening at the vendor level, network operators at every tier need to be paying attention to their supply chain decisions.
What this means in practice
I don't want to end this on a purely cautionary note, because there is a lot of genuinely positive momentum in the industry. The open standards work is real. The telecom-specific AI development is substantive. The collaborative model behind the Open Telco AI Initiative is exactly the kind of industry coordination that produces durable, vendor-neutral infrastructure over time.
For network operators watching these trends, I'd suggest three things to hold in mind:
First, distinguish between AI as a vendor marketing term and AI as an operational tool. The latter is worth paying serious attention to; the former will sort itself out.
Second, watch the Open RAN and Open Telco AI developments. They are early and imperfect, but the trajectory is important. Organisations that build on open infrastructure now will have more flexibility later.
Third, run your supply chain backwards. Know not just who makes your infrastructure, but under what jurisdiction it operates and what legal obligations those vendors carry. In 2026, this is not a theoretical exercise.
We'll keep sharing what we learn at events like MWC as the year progresses. If you want to talk about how any of this affects your network authentication infrastructure specifically, we're ready when you are.
Need help?
InkBridge Networks has been at the forefront of network infrastructure for over two decades, tackling complex challenges across various protocols. Our team of seasoned experts has encountered and solved nearly every conceivable network access issue. For network authentication built on 25 years of expertise rather than statistical guesses, get in touch.
Related Articles
AI intrusion detection
AI can spot network intruders brilliantly, but it can also create more security disasters than it's preventing.
RADIUS design for internet service providers (ISPs)
More than almost any other business, internet service providers (ISPs) need to provide their customers with fast, reliable internet connection to their computer network.