The event, the first in what is hoped to be an ongoing series, was specifically designed to create an atmosphere for interactive discussion and debate between attendees, rather than a series of “broadcast” presentations to a passive audience. To encourage candid input, it was held under the Chatham House Rule, so I can’t name, quote or attribute comments to individual attendees, apart from myself and other speakers.

I gave a keynote to get the debate started, structured around five "provocative" themes and questions intended to stimulate discussion and debate on a number of industry trends and themes impacting the network, some of which are well outside the normal telco focus. Nevertheless, I see all of these altering the landscape for MNOs and other telcos quite profoundly in coming years:

• Does the deployment of 5G SA (standalone) represent a new "Day 1" for 5G? Does it set a baseline for adoption and monetisation of new 5G features and services in later releases of the standard, such as those for IoT, satellite connectivity etc?
  • What new models might arise from "infrastructure convergence" of telecoms and energy? Is there scope to offer services around grid-scale batteries, or perhaps EV charging? Can telcos escape their fear of being categorised alongside utilities?
  • How can telcos shift their mindset and culture more closely to the that espoused in the "Geek Way", a book by Andrew McAfee, which champions speed, openness, science-type trials and iteration, and "ownership" by small teams or individuals?
• Joe Hogan (CTO, Amdocs Networks), who covered a range of other themes from satellite connectivity to network slicing and automation. A key point was whether to view satcos as rivals or partners, and how to deal with uncertainties around regulation, performance and user willingness-to pay. More broadly, this fit with debate on "delayering" of the industry – in some ways, satellites are roughly similar to orbiting 3rd-party cell towers. The slicing discussion sparked debates over how well it works in practice, and the fit with regulatory measures such as Net Neutrality. (My own views on this are well-known and, er, "contrarian" from a telecom perspective).
• Caroline Chappell, Director of CCSquared and a fellow analyst, who looked at progress towards taking full-scale telco cloud visions to market, despite challenges of integration and multi-cloud. There were also roundtable discussions on core networks, network APIs and GenAI for telecoms. There was a very candid discussion on whether the costs and complexities of integrating functions for cloud, alongside the wider transformations needed, outweigh the potential benefits and risks of inaction.
• Amdocs speakers and organisers who kicked off the session and battled against the clock and everyone's vibrant discussion, to keep the discussion roughly on time, included Oleg Volpin, President, Europe, Telefonica Global and Network Offering Division and Itay Gadot, Marketing Strategy Executive.

To be honest, we overestimated the amount of “provocations” needed and the range of separate themes we could cover in a single afternoon, given the vibrancy of discussion – and the quality of the lunch beforehand. It could easily have absorbed two days of time.

Perhaps the most refreshing aspect was the lack of hype, with a great deal of nuance and candour about practical experiences, pointing to the gaps between concepts and reality.

Much of the discussion centred on developments in the cloud, APIs and software-isation of telecom networks, but this was paralleled by another line of thought and discussion around the physical realities of today’s and tomorrow’s networks.

For instance, I covered some of the issues around making 5G work well inside buildings – essential for many of the API and slicing concepts to work in reality. Joe Hogan covered developments in satellite connectivity which highlighted the gap between NTN/terrestrial integration rhetoric and orbital dynamics.

Caroline Chappell’s presentation on network software-isation and telco cloud actually led to some animated discussions about the increasingly diverse hardware choices in networks – less in terms of vendor boxes, but more in terms of underlying semiconductor options for different tasks – will future telcos have to deal with a blend of network systems running on ASICs, Intel, Arm, Nvidia, Broadcom, Qualcomm and other chips?

There was a lot of candour about the realities of cloud-native networks and what that means in practice. There was a general view of the need to push through the pain of integration and adoption, in order to attain the promises of automation and intent, and support a range of 5G’s more advanced features.

One "elephant in the room" was definitely recognised for cloud, even if the solutions were less clear. The telecom industry often exhibits a form of exceptionalism that means that horizontal cloud solutions – especially from hyperscalers – are considered a poor fit for certain requirements… but equally, many smaller operators lack the skills and resources to create and operate their own private clouds. And from the cloud providers’ viewpoint, the relatively small size of the telecoms sector (about 2% of GDP) means it is hard to justify fundamental architectural divergence from their horizontal solutions spanning all other industries.

The paradox is being solved (up to a point) by major vendors packaging cloud elements within pre-integrated solutions, which abstracts some of the complexity, but also reduces the perceived benefits of openness and modularity for 3rd party software elements.

Some of the other learnings and reality-checks included commentary on network slicing, network APIs and the roles of GenAI and the hardware that supports it.

Network APIs came up in discussion multiple times, with considerable emphasis on practicalities currently not addressed by the main industry initiatives. In particular, there was concern over how various "non-functional" aspects would work, such as integration with service catalogues, charging functions and inventory management, plus security aspects such as rate-limiting.

One interesting angle the attendees explored was whether the industry is overlooking the importance of exposure and internal consumption of APIs between different parts of a telco – for instance between the network group and the B2B team. In most other industries, many of the APIs are for the benefit of the IT function itself, and potentially the telecoms industry should focus on similar models.

While I’m a renowned sceptic when it comes to network slicing, it was interesting to hear the discussion that it (today, at least) is being focused more on resource optimisation, such as managing the RAN capacity allocated to fixed-wireless access, rather than pitched directly to enterprises. This is a good example of an internal use of APIs within operators. Other use-cases such as connectivity for broadcast cameras, or applications for public safety and utilities were also proposed. However, the broad consensus was that there are mostly still at demo/trial phase, with various dependencies that still need to be addressed. I also noted that most of the suggestions were for use outdoors.

The debate around GenAI highlighted huge interest in the topic, but also many unknowns, especially given the limited time available. Probably the lowest-hanging fruit is around contact centres and self service agents or bots, where the use of language and text (or verbal) responses plays into the main strengths of LLMs.

There was less consensus on whether or not operators can sell GenAI-powered services themselves – although one strand of conversation suggested that probably B2B use-cases have the most potential, as consumer-grade devices such as smartphones may well “bring their own” GenAI, leaving less scope for telco differentiated services.

Telcos are also going to be constrained in some of their AI-related activities by the availability and economics of the chips needed to run both training and inferencing, as well as the energy supplies needed to run them. While some in the industry have been suggesting that operators may end up providing GPU-aaS, perhaps via edge computing, it was interesting to hear a contrarian view suggesting that some operators will prefer to be buyers of AI compute, rather than owners and sellers.

Overall, the event was a genuinely fascinating and useful one. It involved far more candour and open discussion than is typical. The industry leaders participating were knowledgeable and open to sharing their experiences, and I felt that the industry can still make significant progress if it moves beyond the hype.

That said, it must also adopt more of the geek-way culture shifts towards openness and fast iteration. For instance, in many operators there needs to be a rethink of the slow and hierarchical approach to evaluating new ideas and projects. Decision, feedback and improvement cycles have to shrink – perhaps cutting times by half or even 70%. There also has to be much more openness and clarity – how can developers and customers know what new 5G capabilities to expect, when MNOs don’t even say publicly which 3GPP Release versions or specific features are supported?

One line from the book, a modified version of a well-known Silicon Valley saying, really resonated with me, and fit the tone of the event: "move fast and fix things".