Dean and Ben Talk Tech
Telecom analysts Dean Bubley & Benoit Felten discuss tech news, industry publications, regulation / policy and statistics, relating to fixed and mobile telecom, plus cloud and adjacent domains.
Dean and Ben Talk Tech
DBTT Episode 4: 2026 Preview - Connectivity, Data Infra, AI, Regulation & Geopolitics
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
In this 2026 New Year edition of DBTT we give analysts' takes on what to expect this year, in the areas of:
- Connectivity: What to expect in the evolution path of 6G (and spectrum, standards and AI-Native), FTTH uptake and Copper Switch-Off in Europe, and Satellite constellations, D2D and broadband
- Datacentre Infrastructure: We talk about the likely continued buildout of DCs, with multiple factors influencing location and investment. We talk about shifting dynamics in semiconductors, optical networks and energy supply
- AI Impact: Dean & Ben talk about where AI is having an impact, and why it might take some time to show through in enterprise productivity and economic stats. Also touches on the blurring boundaries of training and inference, and what that might mean for networks
- Regulation: We're anticipating the DNA publication - and wondering when and where it might have effects, and its alignment with other regulatory developments. Also there's Section 230 in the US and a possible redefinition of what it means to be a platform or publisher. And spectrum, again!
- Geopolitics: Sovereignty is certain to be a top theme in 2026, but what does it mean? What are the dynamics to expect between US, Europe, China and elsewhere? Is free speech a defining issue? What about underlying tech stacks? And what else might surprise tech/telecoms on the global stage?
That's a lot to try to fit into 30 minutes, so we don't delve too deeply. We'll pick up on multiple themes in future editions in greater detail, especially after the expected release of the EU DNA on January 20th.
A few links to resources we mentioned :
- This is Dean's analysis of the GSMA spectrum forecast for 2040
- French Competition Authority on Energy/Environmental impact of AI
- The January 15th Unthinkable Lab workshop on AI / APIs
As always, if you like our podcast, please share and give it some visibility, like and subscribe !
Hello. Happy New year to all of you and happy New Year to you, Dean.
dean-bubley_1_01-08-2026_101330and to you.
benoit-felten--he-him-_1_01-08-2026_111325Um, we decided to change our format a little bit for this first, uh, DBTT of 2026. And basically we're going to discuss what we think are key or things that are likely to change or, or be heavily on the agenda. Let's put it that way. In 2026, we've got five different categories, connectivity, data centers, AI regulation, then geopolitics. And, um, maybe we can get started on the connectivity one. What do you have on your list of. Stuff that may happen.
dean-bubley_1_01-08-2026_101330Right. Well, things that are, that are happening is, um, firstly this year is gonna be quite a lot of heavy lifting on defining what six G mobile is gonna be. The standards work is ongoing. Um, and I'm expecting that. To continue at 3G ppp, but also more generally, especially with, you know, angles on, we'll talk about spectrum later. But, uh, things like what does AI native mean for six G? And I'm doing quite a lot of work trying to work out where. We were gonna see automation, whether there's gonna be, um, the connectivity really fundamentally enabled by AI or whether, um, AI is a sort of on the side to improve performance and, and, uh, reliability and, you know, essentially, uh, allow for automation and lower opex. Um, we'll, we'll, I know, I know you've got some thoughts on things like copper switch off for fiber, especially in, uh, you know, uh, for driving fiber uptake, especially in Europe. And the other one we'll come back to is, uh, satellite, but what's, what's, what's, what are you seeing either on the mobile side or on the fiber side?
benoit-felten--he-him-_1_01-08-2026_111325Well on, on the mobile side, the, I think there is a, um, uh, kind of conundrum around the, uh, cost of deploying whatever six, six G is when we finally know what it is. And we've been hearing a lot of pushback from. service providers saying, there's no way we're investing as much on six G as we did on 5G. Uh, and one of the paths to making that investment less burdensome is some forms of infrastructure or spectrum sharing, or both. Um, and I, I am, there's a paradox there because, um, well telcos, especially ex incumbents on notoriously. about sharing anything, but at the same time it might make investment cases a lot more palatable. Um, so I'm, I'm a bit curious as to whether that's gonna become a topic, uh, that, you know, comes to the front a little bit in 2020. So.
dean-bubley_1_01-08-2026_101330It. I mean, in an ideal world, it would've done, I mean, I've been talking about, you know, ideally six G being like spectrum sharing native for quite some time. But to be honest, if I look at what, um, is coming out of, whether it's GSMA where they had their vision 2040 on spectrum recently, which are I, I looked at on my, my substack. Some things that are coming out of the US in terms of US Presidential memorandum on, um, it was like titled Winning the six G Race, but it was mostly about spectrum. Um, and I suspect we may well see something in the DNA in Europe, uh, and certainly 3G PP. There's not much in about the dynamic sharing models and next gen spectrum management. Um, maybe it'll creep in in other ways. At the moment, it seems to be a bit of a last gasp, um, you know, land grab for as much license spectrum as possible, and particularly before the World Radio Congress in 2027. I, I think that the, they're gonna try and keep the message simple of clear as much as you can and, and give exclusive licenses. And I think there's the rest of us around the side saying that's not the most efficient way, but from a conservative point of view and also from the vendor's point of view, it's the most. Attractive way of just sort of selling more radios,
benoit-felten--he-him-_1_01-08-2026_111325Well, unless it pushes everything back by a few more years, so, okay. That, that's, that's interesting. I, I'm, I'm certainly think, I certainly think we're gonna discuss that again this year.
dean-bubley_1_01-08-2026_101330definitely.
benoit-felten--he-him-_1_01-08-2026_111325you asked about cop switch off. Um,
dean-bubley_1_01-08-2026_101330Hmm,
benoit-felten--he-him-_1_01-08-2026_111325paradoxically,
dean-bubley_1_01-08-2026_101330hmm.
benoit-felten--he-him-_1_01-08-2026_111325I don't think. There's much to expect from 2026 on cop switch off. especially because all the regulation you might put in place to ease it or accelerate it or, or facilitate it, you don't have an alternative connectivity solution that's available to most of, um, the subscribers in a given area or even nationally, then you can't switch off the copper. And, and I think certainly in Europe. and I think in a lot of other geographies, most of. The countries that we can have a look at are in that position. So Spain has switched off its copper because it had quasi universal fiber coverage with the last few percentages handled with, uh, primarily fixed wireless access and probably some satellite. Um, France is on its way, and we'll see in a few years how these last few percentage points get handled. I don't see anyone else in Europe that's actually progressing this much. So even though like the commission has said they wanted to really, you know, put tools in place that are gonna accelerate this, I don't see how they do that. So,
dean-bubley_1_01-08-2026_101330Hmm.
benoit-felten--he-him-_1_01-08-2026_111325know. I, I, it's weird because I think it's gonna be a big topic in 2026, but I don't think we're gonna see much change.
dean-bubley_1_01-08-2026_101330I, I mean, my, my understanding was that was one of the grounds for quite a lot of pushback on the draft of the DNA and well, we'll, we'll, we'll talk about that in a, in a few minutes, but, uh, I, I think some markets probably Germany at the top of that list, that, that are, um, not gonna necessarily want to have a really, uh, accelerated, um, timeline for the, for copper switch off.
benoit-felten--he-him-_1_01-08-2026_111325for sure.
dean-bubley_1_01-08-2026_101330Uh, um, and then well, satellite. Um, and, you know, I, I, I, historically, well, like most people, until about three years ago, I'd never really spent as much time on satellite as I did on other forms of wireless, of, you know, cellular and wifi and things. But now satellite is pretty much the top of everyone's agenda. Um. My expectation for this year is, well, the first is we are gonna continue seeing this emphasis on direct to device from satellite. Um, I'm, I think it's gonna be very interesting and it's gonna be very hyped, especially if there's a SpaceX IPO in the middle of it, um, or some other fundraising. But I, I think that, you know, fundamentally it's, it's, it's got hard limitations. I also think we're gonna see a lot of other interesting things going on with satellite, continued adoption of fixed wireless via satellite, um, and some sort of other things around things like earth observation and generally the cost of getting stuff to wall bid is going down. We might see, um, Amazon and others become more competitive with their launch systems. Um, uh, and there's certainly, uh, again, a ton of geopolitics here. Um, and spectrum, uh, issues as well.
benoit-felten--he-him-_1_01-08-2026_111325I, I,
dean-bubley_1_01-08-2026_101330What, what are you seeing?
benoit-felten--he-him-_1_01-08-2026_111325No, no, I think, I think that's right. I, I, I am still somewhat skeptical, uh, with how broad the opportunity is for, for satellite in the connectivity space. Um, and particularly. If things get competitive with Amazon now, you know, full, going ahead. Uh, having launched its first few, um, uh, RAs already, uh, opening service in Europe in the next few weeks, uh, it, it, it is, I, I don't know. It, it, I'm curious to see. If two major American players plus a hypothetical European initiative a few years down the line space to coexist and actually make money on what I still think is a, you know, not insignificant, but not massive subsection of, of the market.
dean-bubley_1_01-08-2026_101330Yeah, I saw a presentation from Amazon Leo, the fact the day they changed the name of it. And, um, that for me makes some sense for sort of B2B use cases because it's, it's clearly gonna plumb directly into the whole sort of AWS dashboard and marketplaces. And so if you're an enterprise and you want to connect a remote site or you wanna backup up, or you know, you want to have workloads which are located or. You know, edge data centers that are next to the satellite ground station. All that stuff, I think makes a lot of sense. Um, but we'll see whether that has an impact on the wider consumer market as well.
benoit-felten--he-him-_1_01-08-2026_111325agree. But e even there, it's gonna be edge cases that, that get,
dean-bubley_1_01-08-2026_101330Hmm.
benoit-felten--he-him-_1_01-08-2026_111325you know, kind of targeted. So you're not, I mean, no, no company in their right minds is going to use that as their primary. Connectivity means for their headquarters, right? So
dean-bubley_1_01-08-2026_101330No.
benoit-felten--he-him-_1_01-08-2026_111325if they have a remote site or two or five, they might use these types of solutions, uh, because it's gonna be cheaper than pulling a fiber there. And it's gonna be better than, than, you know, relying on copper for sure. But
dean-bubley_1_01-08-2026_101330Hmm.
benoit-felten--he-him-_1_01-08-2026_111325having said that, you know, how big is that market? And, and it's, it's really an open question mark to me.
dean-bubley_1_01-08-2026_101330It's, it's gonna be, I think it's certain industries, so, you know, most obviously shipping and transportation and logistics is the most obvious one. But then there's things like oil and gas and mining and resources, you know, for things like, you know, day to day healthcare and stuff. No, because, you know, we know where hospital is. It's gonna have a fiber. But, um, I, I, I will see. Anyway. We should, we should probably move on a little bit.
benoit-felten--he-him-_1_01-08-2026_111325that, on the shipping and logistics, I
dean-bubley_1_01-08-2026_101330Hmm.
benoit-felten--he-him-_1_01-08-2026_111325it's an iot issue more than anything. Like if your, if your device costs, you know, a hundred bucks, you're never, there's
dean-bubley_1_01-08-2026_101330Hmm.
benoit-felten--he-him-_1_01-08-2026_111325economies of scale in making that happen. so that's the real question mark. Okay. Yeah, good point. Alright, so data centers.
dean-bubley_1_01-08-2026_101330Yes. Well, um, well, I, I suppose to, to have a link first off, and I'm, I'm not gonna go through all the multiple reasons. Um, there's a new hype around space-based data centers, and I'm not gonna address the long laundry list of why I think that's over hyped. Um, but it's, uh, you know, cooling and, and latency and everything else, but the main market for. Data center infrastructure. Um, I think this year I don't buy this idea of a complete bubble. I think there's maybe a bubble in valuations, but you, it's pretty clear that NVIDIA is selling all the GPUs it can make, and TSMC, I think the bottleneck is increasingly around the energy supply. And distribution. So, you know, it's things like it's transformers or grid capacity, um, which are the, you know, in many ways slowing things down. So I think there's gonna be a couple of things here. Um, one is we will see continued data center build out, um, and not just in the top tier markets, but in essentially anywhere which has access to power. To cooling and to, you know, favorable site planning. And there's some geopolitics here as well. The things like, uh, subsea cable routes, um, which play into this. Um, I, I think, um, and, and land roads as well. I mean, you're seeing that, you know, people trying to avoid the Red Sea, for example, and go over land to Asia,
benoit-felten--he-him-_1_01-08-2026_111325no, but
dean-bubley_1_01-08-2026_101330through Europe.
benoit-felten--he-him-_1_01-08-2026_111325in country land roads, so you know,
dean-bubley_1_01-08-2026_101330Yeah. Uh.
benoit-felten--he-him-_1_01-08-2026_111325your data centers is gonna be very much determined by, once you've sold the energy bit, how much connectivity you can get.
dean-bubley_1_01-08-2026_101330I, I, I think to some extent, I think the energy is the, the gating factor. I mean, I know I saw presentations from, I think it was Lumen in the US who were saying that they're having to build out fiber into the, you know, middle of Montana or wherever it is, because that's where the data centers are getting built. And I suspect you can probably lay a fiber for a 500 miles faster than you can put in a. You know, high voltage,
benoit-felten--he-him-_1_01-08-2026_111325for
dean-bubley_1_01-08-2026_101330electrical, um, set of pilots and stuff. So I think I agree, but I think, I think that, that, that's the secondary issue of somewhat some of this. What I think we'll see is, well, firstly, if I'm looking at the, the vendor landscape, obviously the semiconductor side of it is growing hugely. I think there is this much talked about shift from. Training to inference, but I think that's a little bit overstated, is now if you look at AI architecture, it's moving towards what's called either continuous or, um, online training. And so I think this sort of binary distinction between training and inference is gonna get, um, uh, you know, fuzzier over time.
benoit-felten--he-him-_1_01-08-2026_111325But
dean-bubley_1_01-08-2026_101330also think we're gonna see a, a shift. Yeah.
benoit-felten--he-him-_1_01-08-2026_111325on that point though.
dean-bubley_1_01-08-2026_101330Hmm.
benoit-felten--he-him-_1_01-08-2026_111325don't you think that, I mean the, there we are in the middle of a very heavy training cycle. Um.
dean-bubley_1_01-08-2026_101330Yeah.
benoit-felten--he-him-_1_01-08-2026_111325Uh, I'm not saying that that's gonna stop entirely, but I guess the shift is going to, uh, become less emphasized on that and more emphasized on usage down the line. My concern is not so much, I guess my concern is as always, like governments in Europe are falling over each other to try and facilitate data center installation, et
dean-bubley_1_01-08-2026_101330Factories. Yeah.
benoit-felten--he-him-_1_01-08-2026_111325by the time they start funding this stuff. They certainly won't be funding, but I don't know if what they will be funding is as sustainable as they think it is, is, I guess is
dean-bubley_1_01-08-2026_101330Hmm,
benoit-felten--he-him-_1_01-08-2026_111325question mark.
dean-bubley_1_01-08-2026_101330uh, uh, it mean, I mean, I've been trying to sort of dig into the, the guts of how a AI architecture is changing. And I must admit, I'm not an expert on this, but the, there seems to be, um, you know, continual or retraining or updating, uh, maybe some of the foundation models will, will. Um, be reduced in terms of sort of what new features we have in chat, GPT six seven, whatever. But I think that, that what we are seeing is a move from just the text tokens to things like voice tokens. So things like voice ai. You're gonna see people doing movement as a native AI input for things like robotics or drones where you actually sort of map the patterns of, of, of things moving around. You really see that in speech. So it may be that we shift from the. You know, consuming all the text on the internet to video, to movement, to sensor input and stuff like that. Um,
benoit-felten--he-him-_1_01-08-2026_111325So we'll be
dean-bubley_1_01-08-2026_101330one thing I think
benoit-felten--he-him-_1_01-08-2026_111325but we'll still need training capability.
dean-bubley_1_01-08-2026_101330you. I think so. Yeah. And I think that the, the, the line between training and inference will blur, I think some of the inference, for example, right down on the end device, you know, if you've got a model on your phone which does image recognition, and you search your photo library for pictures of a, a cat or a, you know, a bottle of wine you, you recognize or something like that. I think that can be static. But I think that some of the other. Domains for AI will need constant retraining. So a lot of the stuff I'm looking at around AI in networks is, you have to think about this with both drift in the model, but there's also drift in the physical environment you will need to retrain because of, you know, new construction in the radio environment. And so what you thought you knew about radio propagation in a cell? Is different in a month's time because there's a new building or construction or road diversion or something. So I think, I think that there will still be training, um, but I think it might look a bit different. And I, I think that then pulls through to the, you know, the ve on the vendor side, obviously the semiconductors. I think, we'll, we'll see some new classes of lower power semiconductor. You saw the acquisition by Nvidia of a company called GR with, with a Kune rather than a K. So it's not the, the same as uh, uh, Elon Musk Grok, uh, recently, and I think we'll see. Headlines, um, about really low power AI using potentially things like neuromorphic and other new radical chip designs. I think that's still mostly r and d, but it's stuff, I think people will start talking about low energy ai and what does that do for infrastructure? Um, one thing I will also say is on the optical side, that's not slowing down. There's a lot of interest in the optical and photonic side of this, and I'm seeing more discussion of things like co packaged optics where you essentially don't have a separate transceiver for the optical chip in the. You know, connecting the processor to the, the network. Um, and that's lower energy consumption and smaller form factor. So I think, I think that optical network domain, you know, you see companies like Nokia. I think, to be honest, we talked about in our last podcast, I think that's a big part of the Nokia Nvidia deal is the optical connectivity.
benoit-felten--he-him-_1_01-08-2026_111325Yeah. Okay. So that sees nicely onto
dean-bubley_1_01-08-2026_101330Hmm.
benoit-felten--he-him-_1_01-08-2026_111325as a topic in and of itself.
dean-bubley_1_01-08-2026_101330Mm-hmm.
benoit-felten--he-him-_1_01-08-2026_111325have the, the, some skeptical views, which as you know, but, um, you've, you're more gung-ho on this, so, so let's, let's start with how do you think
dean-bubley_1_01-08-2026_101330Hmm.
benoit-felten--he-him-_1_01-08-2026_111325this is getting adopted and actually changing stuff? Not. From the provider side, but from the user side.
dean-bubley_1_01-08-2026_101330User side. And, and I hear, you know, I keep seeing people citing an MIT study of, oh, you know, 95% of gen ai, you know, trials and enterprises didn't work. I'm like, well, to be honest, it's very early stage and it was a very small sample. And, you know, the, the trials are so cheap and easy to run that actually, if you can have a one to 20 success rate, that's actually not bad. If you look at it through the lens of a sort of, you know, VC background, I think clearly we're seeing AI. Having more impact in enterprises in, you know, things like software development and, you know, certain areas where you can run agents. So things like contact centers I think are very high on the list. You know, is it gonna essentially reach re cha, you know, change the art the way you produce. Steel or things like that, probably not, but I think it'll crop up in, you know, maybe pharmaceutical, some of the r and d side. And that won't come through in ROI because it's still gotta go through all of the, the rigorous testing and, and multiple phase projects. So, so it, Mel might be the top of the funnel, but the funnel is still five years or 10 years. And so I think it's sometimes very easy to, you know, people say, oh, we, we are running all this stuff and it's not appearing in the bottom line yet. Well. Apart from employment is very hard to sort of generate new additional revenue or improved margin overnight.
benoit-felten--he-him-_1_01-08-2026_111325Absolutely
dean-bubley_1_01-08-2026_101330what the technology is, whether it's AI or private networks or all the stuff we all we've talked about,
benoit-felten--he-him-_1_01-08-2026_111325correct. I
dean-bubley_1_01-08-2026_101330but,
benoit-felten--he-him-_1_01-08-2026_111325that a lot of these trials, I suspect are nonsensical, like, you know, in order to have.
dean-bubley_1_01-08-2026_101330hmm.
benoit-felten--he-him-_1_01-08-2026_111325results, like the first thing you need to do is, for example, not automate something that's not working already, which in many businesses is probably what is happening. rather than say, okay, the process is wrong before we. Try to ai, you know, kind of agent ize, whatever the word, the, the verb is that process, let's actually, you know, assess whether that process is, is working. Um, so, so I think there's probably a lot of, uh. Unintelligent users of ai, which is ironic. I, I guess,
dean-bubley_1_01-08-2026_101330Hmm.
benoit-felten--he-him-_1_01-08-2026_111325so, so I, I would certainly agree with that. I'd add to, to this point, uh, the, I saw a very interesting, interview that was run by, uh, Marcus Weldon, the ex, uh, ex-head
dean-bubley_1_01-08-2026_101330Hmm.
benoit-felten--he-him-_1_01-08-2026_111325um. Of, uh, Nokia's r and d, uh, and Bell Labs, um, is is doing a series of interviews on AI and the impacts of ai, which I, I saw a few of them in a really interesting, but one of them was talking to an economist about jobs and, and the interesting thing is that, um, they are absolutely seeing effects on employment already. So, you know,
dean-bubley_1_01-08-2026_101330Hmm.
benoit-felten--he-him-_1_01-08-2026_111325it's not a, It's not, it's, it's, it's not a scare. It's actually happening, but it's happening in very different ways on very different jobs. and what's interesting to me is that the jobs that seem to be impacted are kind of middle class jobs. They're not. they're not,
dean-bubley_1_01-08-2026_101330Yeah.
benoit-felten--he-him-_1_01-08-2026_111325physical jobs for the most part, and they're jobs that previous industrial shifts not really impacted negatively
dean-bubley_1_01-08-2026_101330Hmm.
benoit-felten--he-him-_1_01-08-2026_111325rather impacted positively. Um, so I'm actually curious to see when, because I think it's inevitable when that turns into a political pushback and how, you know, unfortunately. When the lower class is impacted, there's a usually a big consensus to say, yeah, you know, it'll sort itself out. No, no worries. But it's when it's your core electorate, the people who
dean-bubley_1_01-08-2026_101330Hmm.
benoit-felten--he-him-_1_01-08-2026_111325go out to vote that are impacted. I think one of the things we might start to see in 2026 is actually this coming into the political discourse in a much more forceful way. Not necessarily effectively, but I do think we're gonna see that happen.
dean-bubley_1_01-08-2026_101330Uh, we can see protest marches of accountants and people who used to work for McKinsey in Deloitte. Yeah, I maybe.
benoit-felten--he-him-_1_01-08-2026_111325early for that, but, but I think we're gonna see some politicians and some parties actually take that on as. You know, a problem that needs to be regulated at the very least or sorted.
dean-bubley_1_01-08-2026_101330I, I, I think, I think there's that, where I think that's also gonna get accelerated, potentially dunno, it's this year or, or another year or two, is around what's been called physical ai, which is, which is also an opportunity for, you know, the telecoms industry, which is connectivity for robots. So in, I, I was very skeptical initially, or the idea that a humanoid robot was what we needed. But actually, given that most business processes, physical ones are designed around, you know, a human of, you know. Yeah, whatever rough, normal human dimensions are having, um, the ability to load a truck or to, um, push a trolley around a hospital or deliver food or something like that. I think that's perhaps also where there's gonna be some impact from an employment point of view. And I see some. Implausible, implausible forecast from, again, from sort of musk with, I've forgotten the name of Tesla's robot. Is it Optimus? Or something like that. Um, and there's, there's about, you know, 20 other different, I'm sure as recording, CES is on in, in Las Vegas. And I, I, I expect to see the sort of robotics videos. Coming out thick, uh, you know, fast and furiously over the next few days. Um, so I think that's an area one that I'm looking at is around, I've already mentioned the contact centers is voice and ai and some of that's gonna be useful and that's not gonna be replacing humans. It's gonna have a voice agent that screens your calls for you and looks for. DeepFakes and scams. But certainly, you know, you wouldn't want to be a, a, a contact center operator in, you know, over the next three to five years, um, when it's gonna be possible to have real time speech in multi multiple languages with all your sentiment analysis and access to all of your customer data and CRM systems. Are we gonna have human, human contact center agents? Well, a very small number, I think.
benoit-felten--he-him-_1_01-08-2026_111325Well, the other thing is no one wants to talk to contact centers when they're pushing it anymore, and the companies don't want you to talk to people anyway, so, yeah. see
dean-bubley_1_01-08-2026_101330Yeah. Yeah.
benoit-felten--he-him-_1_01-08-2026_111325one last thing on this. Uh, next week I'm in Brussels for an, uh, open internet event that AEP is organizing to
dean-bubley_1_01-08-2026_101330Hmm.
benoit-felten--he-him-_1_01-08-2026_111325the 10 years of the open internet regulation. And at that event, they're gonna present a study that they published in December on AI and energy. Um. I have deliberately not read it because I'd like to hear how they talk about it before I actually read the documents. Uh, but that's maybe something we'll discuss, uh, next time.
dean-bubley_1_01-08-2026_101330Uh, definitely. And also next week I'm running in London the my next Unthinkable lab event, which is looking at, um, telecom APIs and whether AI ends an AI agents end up being what consumes telecom APIs rather than human engineers and developers. Um, and this particular technology called MCP that might help there, but I'll, I'll report on that next time we speak.
benoit-felten--he-him-_1_01-08-2026_111325Okay, cool. Alright, so let's move on to the last section, which is regulation and geopolitics. So obviously, uh, everyone has, uh, on their agenda right now, January 20th, which is when the European Commission, uh, will supposedly, uh, publish the DNA, um.
dean-bubley_1_01-08-2026_101330A long awaited DNA. Yeah.
benoit-felten--he-him-_1_01-08-2026_111325way to DNA. Yeah, absolutely. The, the weird thing is that as important as this is going to be in all the discussions in 2026, I don't think whatever is in there is gonna have an impact for quite a while still, because the parliamentary process and the whole political process within the European institutions takes a long time. So unless. Unless, well, in fact, even if there is not much that changes in there, pa some parliamentary are gonna try to push for more radical changes. Others are gonna try to fight back. So it's, it's really hard to think that this is gonna have an impact at the very least, 20, 27 and possibly
dean-bubley_1_01-08-2026_101330Mm-hmm.
benoit-felten--he-him-_1_01-08-2026_111325Now, you had a point about this also, which I think is interesting. It might be a bit more short term about how this overlaps with other. of
dean-bubley_1_01-08-2026_101330uh. I think one of the things I've noted is, is divergence or even orthogonality between different regulatory focus areas, particularly in Brussels. And so I'm gonna be looking to see what's in the DNA and does it roughly align or does it work against what other things are being said in, in the, by the commission on, um, data, on ai, on chips, on cloud, on, um, you know, privacy on, uh, energy. As well come to that because you know, all of those, you know, in theory all ought to sort of tie together. But what I've noticed in the past is, is that they tend to sort of, when you get to other detail, you suddenly realize that on one hand Europe's talking about the sovereign cloud industry. And on the other hand it's trying to think about regulating IP interconnect or something when you actually need more interconnect and a dense mesh of it. And there's a few other sort of. Um, uh, examples of, of inherent contradiction and hopefully the DNA has addressed that and ends up being more in alignment with the other goals rather than, and trying to take off part of the industry in the wrong direction.
benoit-felten--he-him-_1_01-08-2026_111325Well,
dean-bubley_1_01-08-2026_101330I,
benoit-felten--he-him-_1_01-08-2026_111325from our next installment onwards that that's one of the things we're gonna have to look into for sure. One
dean-bubley_1_01-08-2026_101330I mean, so all the things that.
benoit-felten--he-him-_1_01-08-2026_111325yeah, sorry. Go on.
dean-bubley_1_01-08-2026_101330Yeah, I was gonna say also, you know, in terms of a regulation point of view, other stuff I'm looking out for this year is in the UK there's Ofcom doing the, the telecom access review, looking at things like wholesale for fixed networks and, and where Openreach sits in that. There's gonna be a bunch of stuff on the spectrum side around. Um, you know, preparations for World Radio Congress 27, which comes at the end of 27, but it's like a four year cycle. And this year there's a lot of the hard work and there's a, a continual stream of stuff coming out in the us whether it's on wireless, on spectrum, on, um, the funding for connectivity in rural areas and satellite and everything else. So, uh, I'm sure we will, we will have plenty of, uh, things come out of FCC and NTIA to, to look at over the next 12 months.
benoit-felten--he-him-_1_01-08-2026_111325One thing that. Uh, I, I am wondering about is like in the last few years, but it's, it's picked up again in the last few weeks. We've, uh, we've heard discussions in the US about repealing, uh, section two 30, which protects social networks and platforms. from being litigated on stuff that is posted on said platforms. and it's been a, a, an underpinning of allowing the internet in general and platforms in particular content without that risk of litigation. we have similar regulation in Europe. I don't want, I don't actually know what it's called, but, but, uh, you know, I know that there's similar things in Europe. Um, what, what. What I think is interesting here is that, um, with all this talk about free speech, which is gonna see into our next, uh, our next segment, but, uh, there is a fundamental question about. Accountability of the algorithmic nature of the platforms. In other words, it's not because something is posted that they might be responsible, it's because something is pushed. And also, um, the, the other aspect, and this has been highlighted by the frankly scandalous, um, strip, a photo feature that emerged on XA couple of weeks
dean-bubley_1_01-08-2026_101330Hmm.
benoit-felten--he-him-_1_01-08-2026_111325is. The Section two 30 specifically is designed to protect the platforms from stuff that is posted by people. Um, and there is a very real argument that this is not stuff posted by people. and so
dean-bubley_1_01-08-2026_101330Hmm.
benoit-felten--he-him-_1_01-08-2026_111325am wondering if we might come to some reconsideration of some of the key underpinnings that currently protect. A lot of the tech industry from any forms of liability on the content that gets, um, carried
dean-bubley_1_01-08-2026_101330Uh,
benoit-felten--he-him-_1_01-08-2026_111325their networks.
dean-bubley_1_01-08-2026_101330I, I, I think you might be right, and especially the distinction between platform and model. From an AI point of view is gonna become very, very blurred and it already is with, you know, X and, um, GR for example, or for that matter, meta and, you know, a bunch of others. I mean, you know, chat, chat, PT and open AI and perplexity are going the other direction. Um, you know, I noticed even my browser's got a, usually five has got a perplexity, the icon on it if I want to use it. And obviously you've got copilots everywhere and, um, Google, Gemini. So maybe the, there's gonna have to be some. Thought about, well, I, you know, what you do when you say, well, I didn't post it, my AI did. Or your AI as a platform tuned it and tweaked it. And so I think this certainly will need to be, you know, here it's gonna depend on the way you are in the world, either case law or some much better definitions.'cause I'm, I'm not sure, I don't know the wording of section two 30, but you know, clearly it was done in a, an earlier era and we're now in a very, very different, um, landscape.
benoit-felten--he-him-_1_01-08-2026_111325absolutely true. So we move on to the broader sovereignty discussion, right? Free speech is becoming a political, uh,
dean-bubley_1_01-08-2026_101330Yeah.
benoit-felten--he-him-_1_01-08-2026_111325dispute point between the
dean-bubley_1_01-08-2026_101330Yeah.
benoit-felten--he-him-_1_01-08-2026_111325and Europe for sure, but obviously. China has a different view on free
dean-bubley_1_01-08-2026_101330It
benoit-felten--he-him-_1_01-08-2026_111325as well, so,
dean-bubley_1_01-08-2026_101330yes. Yeah.
benoit-felten--he-him-_1_01-08-2026_111325you, I mean, how do you see that impacting stuff in
dean-bubley_1_01-08-2026_101330Uh, well, I mean, the most obvious one is the US is really pushing hard against, um, digital services, rules and taxes in Europe, uh, whether that's eu, uk, or elsewhere. Um, and so the sort of DSA and to a lesser extent the DMA, uh, so, so for, that's the Digital Services Act and the Digital Markets Act. I sometimes get. Conflated, uh, and there's various things around privacy as well. Um, and so, you know, you, you do have this sort of us trying to use the sort of. Free, free speech absolutist position, um, against some of the fundamental underpinnings of, uh, particularly EU regulation on tech and telecoms. Um, I'm not sure what that's gonna look like. I mean, yeah, there, there's a, there's a, a bleed across here into the wider discussions around sovereignty. And when you get into sovereignty, there's, I think we talked about on the last day, there's sort of data sovereignty, there's policy sovereignty. There's. There's the actual tech stack and supply chain, lots of different angles here. And so clearly, you know, this is gonna be a pushback on historically, frankly, Europe has always been quite proud of exporting its regulation, whether it's GDPR or anything else. Uh, and I think it may end up in a world where it's got some imports as well.
benoit-felten--he-him-_1_01-08-2026_111325Well, I, I dunno. I mean, there's two ways this can play out, right? And I'm gonna put the threat of military intervention to the side for a second, even though I'm not sure we can do that anymore. But the, the, there's a point at which what you lose if you go full ahead on what they don't want the us, Becomes the only viable option. In other words, once the threat has been implemented anyway, which the Trump administration has done with tariffs and is doing now with threatening politicians, et cetera, and the way it would with, you know, a dictatorship that they didn't like, there's a point where actually doing what they don't want you to do doesn't have any. Additional associated cost anymore. the thing that I'm wondering about is, mean, it's interesting to see that with the Venezuela thing, a lot of, very sovereignty is high on the agenda. Politicians in Europe are actually not so in love with Trump anymore. Or at least they can't publicly say that they are because, well. Obviously invading a, a neighboring country is not, uh, respectful of sovereignty. And, and
dean-bubley_1_01-08-2026_101330yeah.
benoit-felten--he-him-_1_01-08-2026_111325thing that I'm wondering about is, is there at some point a political consensus to say, well, actually, this is now number one on the agenda and we're gonna go full ahead on it, and there's gonna be consequences, but we're already living through most of these consequences anyway.
dean-bubley_1_01-08-2026_101330Uh, I mean, I, I think, I think the sovereignty is gonna be the buzzword of the year for whether it's telecoms or tech generally. Um, I'm, I'm probably gonna be doing an event in, in April on that, and there's lots of different. You know, touch points and the question of how sovereign can you be? I mean, you know, maybe French Misra AI company will build a data center in Greenland, uh, or something like that to you. But, uh, uh, which, uh, you know, we might end up debating, but, uh, yeah, I think that, that, on a more serious note, I think that that the definitions of sovereignty. We'll become clearer this year, the moment people throw this term around. And as I've said, there's multiple layers and abstractions to it. So I think that, you know, we we're gonna, and I also think there's gonna be some pragmatism. I think what I do think is that, um, Europe may have overreached with, with some of its extra territorial objectives. On things like GDPR and I think that the US is possibly overreaching with its ter extraterritorial objectives around free speech. Um, and clearly it's not gonna get much, um, uh, traction in China and other places. Um, so we'll see. To, to me that seems to be amenable to, you know. Hardball diplomacy, you know, in much the same way that defenses and I, I don't think that there's gonna be anyone pulling the plug on Azure, um, or AWS in Europe. And there's no one gonna be pulling the plug on, you know, A SML or ARM in the us. And I think that there's gonna be some accommodations that will be made, probably there's gonna be an ebb and flow and we have to deal with. Whether it's White House or Brussels or Westminster, making inflammatory statements occasionally. Um, but I think that's part of the, the, the landscape is you have to deal with, uh, occasional blowups.
benoit-felten--he-him-_1_01-08-2026_111325the one thing that I think is, uh, maybe worth considering and would have some impact for some of our listeners is that I think there's a very fundamental difference between US-based infrastructure services that Europe cannot do without, at least in the short, midterm, most likely in the long term. Um, and
dean-bubley_1_01-08-2026_101330Hmm
benoit-felten--he-him-_1_01-08-2026_111325let's call them entertainment services that. If they were to be targeted, it might be an outrage from a consumer point of view, but it would really impact
dean-bubley_1_01-08-2026_101330mm-hmm.
benoit-felten--he-him-_1_01-08-2026_111325to put it another way. Netflix is very easily substitutable, uh, whereas, uh, AWS is not,
dean-bubley_1_01-08-2026_101330Yes.
benoit-felten--he-him-_1_01-08-2026_111325and
dean-bubley_1_01-08-2026_101330Um, and I think, and I think chat, GPT is somewhere in the middle. If you've, if your kid's got their homework projects in chat, GPT or something like you or your enterprise uses the API that's looking like infrastructure.
benoit-felten--he-him-_1_01-08-2026_111325It, it is. But there are a lot of competitors that you can relatively easily switch to with a bit of relearning, uh, which
dean-bubley_1_01-08-2026_101330Ah, yeah.
benoit-felten--he-him-_1_01-08-2026_111325which is not true of, of, you know,
dean-bubley_1_01-08-2026_101330Hmm.
benoit-felten--he-him-_1_01-08-2026_111325or physical infrastructure. All I'm saying is I would not be overly surprised if some regulators started looking at this and saying, well. I don't have any qualms about targeting certain services anymore because I know there's no systemic impact. You see what I mean?
dean-bubley_1_01-08-2026_101330Hmm. Yeah. Yeah. Well that's, that's fine. Until Netflix buys, um,
benoit-felten--he-him-_1_01-08-2026_111325Oh,
dean-bubley_1_01-08-2026_101330or Paramount buys. Yeah. Whoever. Yeah. Yeah, yeah,
benoit-felten--he-him-_1_01-08-2026_111325but you know
dean-bubley_1_01-08-2026_101330No, yeah.
benoit-felten--he-him-_1_01-08-2026_111325I didn't, I, the idea was not to target Netflix specifically, but it's more to use an example of what's primarily an entertainment service as opposed to. And, and
dean-bubley_1_01-08-2026_101330yeah, and clearly, clearly, well, although interesting there it becomes very hard to say, oh, we're gonna block X without saying, we're also gonna block TikTok. And so you end up with a bit of a multi-way mutually issued destruction with not just us, but probably China as well.
benoit-felten--he-him-_1_01-08-2026_111325Yeah. Well, that's a bleak note
dean-bubley_1_01-08-2026_101330Ah, that's a cheery one anyway, but, uh,
benoit-felten--he-him-_1_01-08-2026_111325year
dean-bubley_1_01-08-2026_101330I would actually, actually, one last one on geopolitics is to watch out for things which are outside of our day to day. And one thing I noticed was there's a diplomatic pac between China and Japan at the moment over, um, you know, various things, particularly Taiwan. Um, but uh, China is, is putting an export ban on not just things like, you know, automotive and rare earth, but some of those things like. Um, chips and, uh, compute network equipment. And so we might have to also think about the sort of, you know what as well as the sort of Europe US is, is what the Europe, China, uh, dynamic looks like as well. And, and, and also secondary effects from Japan and South Korea.
benoit-felten--he-him-_1_01-08-2026_111325Absolutely. Yeah. Alright, well, um, that
dean-bubley_1_01-08-2026_101330We, we come?
benoit-felten--he-him-_1_01-08-2026_111325one. Very,
dean-bubley_1_01-08-2026_101330Yeah.
benoit-felten--he-him-_1_01-08-2026_111325we ran over a little bit. Hopefully we didn't lose too many of you. But, um, to discuss in 2026 for sure. Thanks for
dean-bubley_1_01-08-2026_101330Yeah. And, uh, thanks Benmar. And, uh, yeah, feel free to, well hit, hit subscribe and reach out to either of us if you'd like to do a deep dive on any of these things.
benoit-felten--he-him-_1_01-08-2026_111325Absolutely. Great. See you
dean-bubley_1_01-08-2026_101330I.