This year’s iteration of Future Net World Conference was buzzing with excitement. The discussion halls were busier than ever, and the mood was ebullient. The hubbub was down to the inclusion of AI as one of the conference’s key themes, alongside Future Net’s usual coverage of OSS and BSS. Of course, every telecom event in the last three years has focused heavily on AI, but it is finally starting to have a meaningful impact on the, at times, dry area of OSS/BSS.
This is not to say that every operator present was running TM Forum Level 5 autonomous, self-healing networks. Of course not; this is Europe.
But there was plenty of excitement as to how AI was being implemented in operations. AI has long been at work in network operations, many years before the arrival of generative AI, but there is now a lot more momentum and collaboration, pushing things forward at pace.
The broader themes of ‘Network for AI and AI for the Network’ were highlighted in the keynote panel, a discussion between the CTO’s of some of Europe’s largest telcos – Orange, Telefonica, and Vodafone, alongside networking vendor Ciena.
The discussion of these broader themes showed a hesitancy to commit to some of the most high-profile AI projects in the sector at the moment, such as edge and sovereign services. However, there was strong interest in AI for networks, to improve autonomy and efficiency.
The discussion began with the impact AI was having on network traffic, which, according to these operators, was minimal. You will notice expressions of tentativeness (“it is expected,” “we are waiting on the evidence,” “pretty sure”) in the statements below.
“We all know that we are going to have an increase in the uplink content but the truth is that today, this is not something we are seeing in the network,” said Andrea Folgueiras, CTIO, Telefónica Group. “It is expected to double but we are monitoring this all the time and we are not seeing this today.”
“The evidence today is pretty anemic,” said Scott Petty, CTO, Vodafone Group. “The uplink growth will be driven by massive content creation, maybe smart glasses. That’s the theory; we are waiting on the evidence.”
“We are watching the traffic very closely. What we see is that for some applications, the uplink is becoming more important. It is still very small, but it will be important, I’m pretty sure,” said Laurent Leboucher, EVP Networks & Group CTO, Orange.
The panel moderator, Mark Newman, Chief Analyst at TM Forum, pointed out that the overall growth of traffic volumes has been slowing in recent years, which means that there is more available capacity, if networks are designed around the current growth models.
Edge services
The next big question about AI is whether the expected demands on the network will require far-edge computation services. Given the lack of evidence of an AI boom in the quotes above and the general lack of impetus from European operators to get on board with the AI-RAN pitch espoused by some in the US and Asia, it is no surprise that the operators were not, by and large, sold on the market for far-edge services.
The strongest enthusiasm for edge services came from Telefonica’s Folgueiras, who described how the telco has launched edge services for B2B customers.
“We believe that this is a key element of the future,” said Folgueiras. “We are deploying 17 edge nodes in Spain, and we are starting to work with that in other geographies. We are already talking with CAF, the train company, we are trialing a video system that can monitor the trains’ security.”
At MWC this year, Deutsche Telekom, Orange, Telefónica, TIM, and Vodafone announced a pan‑European edge federation, essentially building an interoperable edge cloud that spans their home nations.
These examples are evidence of a new commercial opportunity for telcos, but it is worth remembering that 17 new nodes across the whole of Spain is a very different proposition than a remastering of an entire RAN network, a la the AI-RAN pitch.
Spain has around 40,000 tower sites. There are currently 811,237 towers in Europe, a figure which is expected to rise to 858,687 in 2031, according to a recent forecast from our sister unit RAN Research, see separate article this week.
What was being pitched was a conservative imagining of the “network for AI,” and one that has been overheard from European operators in recent years.
Low interest in sovereign AI
The next biggest opportunity for all MNOs, and especially in Europe, is talk of sovereign networks and services. When asked how real this proposition was, the panelists were surprisingly lukewarm on the topic.
“Sovereign cloud or sovereign AI without a sovereign network does not make sense, because you’re connecting an important set of data applications on the endpoint that you would not get control over. The opportunity for us as telecommunications providers is to provide certainty around how the network is operated, where it is operated, who can access the data, and how the data is encrypted. This is an opportunity for B2B services, I think this is where the opportunities are more than sovereign,” said Vodafone’s Petty.
“Whether people are willing to pay the incrementally higher cost for sovereign services is not clear yet. We will understand more about this in the coming months,” Petty noted.
Orange’s Leboucher displayed more enthusiasm for sovereignty, but his description of sovereignty was quite broad.
It involved a certain type of regulatory overview, which sounds like an easier setup than building a network made from equipment that is sovereign to a certain country.
“If I look at the latency between two large cities, Paris and Marseille, with optical fiber, we can get five milliseconds. So, it is extremely hard to monetize new cases where you really need the very far edge. If I put that aside, it’s more about providing capacity, which is sovereign in the sense that it is controlled according to the regulation of a certain government for some specific, very critical workloads. And second, it’s more about trust – how you provide a sufficiently trusted environment for your customer, for your enterprise customers,” said Leboucher. “For instance, selecting the routes so that the data avoids some countries for geopolitical reasons.”
In summary, there was a wait-and-see approach to the sovereign idea. “The regulations are not very clear. A lot of work on sovereignty mechanisms is needed. I think, market-to-market, it will become more clear in the coming quarters. What’s most important to us is how we harness AI today, for ourselves. How do we become world-class at deploying AI,” stressed Petty.
Petty may have been referring to Vodafone’s shared services arm VOIS, which is responsible for managing the adoption of AI for areas such as operations, finances, and customer services. Wireless Watch spoke with the Director of Commercial & Go-To-Market at VOIS last month about activity in this department.
None of the operators on the panel were building a sovereign AI factory. To look around at the current sovereign AI activity in Europe, there is a sense that current services are not doing especially well.
Deutsche Telekom, a vanguard in this space, is almost constantly campaigning for more customers for its sovereign AI factory in Berlin. At MWC this year, Deutsche Telekom CEO, Tim Höttges, called on the German state to start using DT’s sovereign data centers, which implied that it was not yet doing so.
Telenor Norway began building an AI factory in 2024, and while it says that this product launch is going well, Telenor execs have described a cautious approach to the enterprise.
Telenor’s Group CTO, Cathal Kennedy, described how the project started small, as a minimum viable product (MVP) in an interview with TelecomTV, to avoid the risk of overbuilding.
“We’re not a company that has bought 10,000 GPUs; we’re buying hundreds of GPUs, so we are scaling with our customers,” said Kennedy. “Now, two years in, we’ve got enough customers for us to actually be satisfied with, and we’ve built the capability to put our own workloads for our own network development. So, one thing is, you can buy a GPU and put it in your data center, but it is another thing to be able to utilize this,” said Kennedy.
Autonomous networks
You might think this discussion would lead to a total rejection of all new business opportunities for the sector, but there is a single, unified bright spot, and that was the use of AI for networks. There was a strong interest in this subject across the board.
“The single most powerful driver right now that we’re seeing in the AI investment side has to do with the financial performance. Everyone has wanted to make these changes for years, but hasn’t been able to, and that’s cleaning up the 50-80 applications they may have in the OSS that are causing operational churn. And collapsing those applications down in a way that allows them to have a clean data set,” said Joe Cumello, Senior Vice President, GM, Blue Planet, a division of Ciena.
“We cannot let our technical debt accrue again, where you are spending 80-90% just running the things you have built in the past,” said Vodafone’s Petty. “That’s why this message is so important; how you build your infrastructure is so important. It frees up the oxygen to run the network.”
“The journey we are going on with autonomous networks is allowing us to plan in a couple of days instead of months.
The goal is not to reduce tickets; it is to try to avoid tickets. This is a huge space we are working on,” said Telefonica’s Folgueiras.