The end of big capex in telecoms: causes and implications

Caroline Gabriel

Welcome to the Analysys Mason podcast. My name is Caroline Gabriel and I'm a Research Director heading up our work in telecoms networks and cloud. I'm here with Rupert Wood, a Research Director who leads, among other things, our Operator Spending programme and is an expert in telecoms industry capex trends.

Today, we'll be discussing a foundational question for the whole telecoms industry: whether we have seen the end of huge capital investment projects and high peaks of telecom operator capex and are entering a period of steadier and overall lower capital spending.

Rupert has recently published a major report entitled ‘The end of big capex: new strategic options for the telecoms industry’, which will be the basis of our discussion today. An associated article is available to access and download from the Analysys Mason website.

So Rupert, in your report, you say we're entering an era of plentiful, ubiquitous and affordable bandwidth, but you also see the sure signs of a crisis of overproduction. Can you describe some of that and what you see as the broad trends in capex now?

Rupert Wood

Yes. So I think there was essentially a convergence of two major trends peaking and that's led to a very high capex intensity. I think they peaked in about 2022 and are now clearly falling. Those two big trends that have converged are fibre build, which appears to have peaked, and the initial 5G build.

Fibre build has been the dominant aspect of capex, particularly in Europe, for the last few years. But fibre build has a one-off quality; you build it once and you can leave it – it has a very long, useful asset life. Whereas 5G is the last of several generational cycles of spend.

So when we look at the 50 largest spending operators that have capex in excess of USD1 billion a year, the trend is down pretty well everywhere, mitigated somewhat by some East Asia spending on businesses adjacent to telecoms.

Caroline Gabriel

Are there any other countertrends that you can see?

Rupert Wood

Well, not many. In developing markets, there's still some growth, in particular in the lowest-income areas. A couple of areas that I think are countertrends: fibre-poor operators, including many cablecos, still face some fairly stark spending choices, whether to self-overbuild with fibre or spend a lot of money on DOCSIS 3.1 or DOCSIS 4.0. I think the USA fixed broadband, looks strong for the time being, especially with all that BEAD money coming in.

Caroline Gabriel

But are we also not seeing spend shifting from larger operating groups like you're talking about to smaller ones, such as all those altnets, and also from operators to specialist infrastructure businesses, so there's capex coming from other sources? 

Rupert Wood

Yes. I think those are both true. We do see a shift from the larger players in fixed line, particularly towards a large number of smaller altnets. And yes, it's certainly true that spending capex has, to some extent, shifted from integrated operators to specialist infrastructure businesses like TowerCos or FibreCos. But in both those cases, I think capex is peaking. Carve-outs can help the marginal FTTP business case, and neutral host players can do that, too. But netcos, ultimately, operate in the same demand environment as traditionally structured operators.

Demand dynamics in telecoms

Caroline Gabriel

So what about that demand?

Rupert Wood

Well, it's weaker than it's ever been. Data traffic is only one way of many to gauge demand, but for better or worse, it remains pretty central to decision-making amongst operators.

Operators could define their mission in different ways, but I think at the core of every operator's business is the production and sale of bandwidth. That's essentially what's at the centre of it all: capacity, which is filled or stretched by the volume of traffic and by peak demands on bandwidth. On fixed, we see growth in the teens at best. I think our last analysis of Europe suggested that in 2024, growth would be about 12%, this is in part driven by still-rising broadband connections.

In mobile traffic, it's important to split out fixed-wireless access here, but growth is, in some countries, falling to single digits per annum. In Europe, we see stronger growth than in many other places. 2024, we're estimating a 15% growth in mobile traffic. In some of the more advanced countries, on a per-subscriber basis, that growth has actually fallen to zero. The reason I mentioned fixed-wireless access is that CTIA reported a 36% growth in mobile network traffic in 2023. But if you split out fixed-wireless access, which depends on valid capacity on 5G, then mobile device traffic growth is single digits. The same is true for advanced mobile countries like Japan, Korea, Finland and China.

But what I want to emphasise in all of this about demand is that the dynamics of IP traffic growth are now almost entirely driven by fixed broadband and not by demand intrinsic to mobile. That's not only because fixed broadband traffic remains globally about six times higher than mobile, but it's also because the biggest determining factor of both the volume and the growth rate of mobile network traffic is the absence of, or unaffordability of, fixed broadband, especially fibre. It has very little to do with anything intrinsic to mobile networks itself, the mobile ecosystem of mobile applications and mobile devices, which all work as well on fixed Wi-Fi.

So as wireline penetration increases and the number of people that are dependent only on mobile for connectivity increases, then fixed line will squeeze mobile demand rather than the other way around. So, in a way, looking for what drives traffic on mobile is a wasted effort.

The role of new applications

Caroline Gabriel

Do you think new applications will come along to change this trend?

Rupert Wood

Yes, sure, they will. But by that stage, and already, the headroom has become much larger. It's difficult to see anything now that would drive a real spike in investment in the capex-intensive access parts of the networks. We've thought about AI, and we believe that that has the potential to increase traffic dramatically on the long-distance or the metro parts of the network, but not in the access part particularly. XR might still drive something or other, connected vehicles, that may happen, but as I said the headroom has just got much bigger than it's ever been.

Caroline Gabriel

You described something that's looking like a collective hallucination in the industry about the relationship between demand and capex. Can you flesh that out?

Rupert Wood

Yes. A lot of people talk about demand, particularly in terms of data traffic or bandwidth, as driving capex, and that can be a bit of a myth in some ways.

First of all, if it is true, it's probably more true on the mobile side than on fixed, that relationship between a large amount of demand and having to spend more to serve it. I'm not convinced it's entirely true. But, the point is mobile isn't actually such a large part of telecoms investment as many might assume. It's probably only about 25% of capex in Europe. 5G has been about 20% of capex in Europe. Maybe it's a bit more globally. So it's not representative of the whole operator business.

Another thing we often see is CAGRs, which are a sort of widely abused measure in the industry. We tend to extrapolate from current growth rates, but actually, those growth rates are in constant decline. So, CAGRs are pretty useless for describing the top end of an S-curve.

Another fact is that quite often, you hear, particularly from vendors, confusion about the pattern of replacement for a pattern of demand. Patterns of replacement, the timetable of replacing parts, is driven by useful asset lives and by that stage, something newer and faster will come along, and productive forces drive unit costs of higher capacity equipment down to near the levels of the original equipment. So, in fibre, you go from G-PON to XGS-PON, and the unit costs are probably pretty similar after 8 or 9 years. But that isn't a pattern of demand, that's just a pattern of replacement.

Industry lobbyists are pushing the idea of extracting traffic fees or similar from content and app providers, so that's another reason that there's perhaps too much emphasis on the relationship between traffic, demand and capex.

But I think there's another reason: people quite often misunderstand or misrepresent the relationship between applications and devices, and the developers of applications and devices, and networks. Devices and applications are generally developed to fit networks rather than to stretch them. There's a tendency to overstate the convergence, the integration of IT and telecoms. Yes, that's possible, but it's not obvious it will make a big enough change to make a great difference. So, the world of application development and the world of device development, to a large extent, sits on a utility telecoms layer on an internet layer rather than on a transactional mobile layer, which it's usually couched in terms of mobile networks. So, it's that relationship between applications, devices and networks which I think is sometimes misrepresented.

The crisis of overproduction explained

Caroline Gabriel

So, what do you think are the symptoms of this hallucination?

Rupert Wood

I think it's a crisis of overproduction. I mean it in a quite standard Marxist sense. Traditional productive forces drive towards something, irrespective of actual market demand. But this crisis of overproduction, of producing too much bandwidth, too much equipment which offers too much bandwidth, is often presented as a crisis of under-consumption, and people tend to point to a nebulous latent demand rather than real demand which miraculously corresponds to the expansion of capacity brought about by the new equipment.

Then that leads to the fact you've got too much capacity in networks, and it can't be sold tends to generate a sort of Micawberish belief that something will come along, some new service or new demand will come along and fill the networks, fill the pipe.

Another symptom of this is simply piling it high, and selling it cheap. Not just unlimited data plans which are pretty common now, but even further, fixed-wireless access is the ultimate pile-it-high, sell-it-cheap service. But then there's a counterpoint to that as well which is a belief in or a collective struggle to create or substantiate more indirect business models of ways of selling networks to content application providers. For example, these B2B2X models, not just wholesale, and hope somebody else will take the responsibility of actually selling the capacity to an end user who wants it.

Caroline Gabriel

But is this capacity headroom anything new? I mean, surely that's what operators have always been doing, putting in capacity well ahead of demand?

Rupert Wood

Yes, of course, they have, and rightly so. But I think it's a matter of scale in some ways, and it's a matter of where we conceive demand is heading. I think the industry has never been so far ahead of demand as it is now.

Putting in faster networks is driven mostly by competition rather than by inherent demand for that kind of bandwidth. That competition is sometimes artificially imposed in a highly regulated environment like Europe, and it's driven by the Net Promoter Score associated with higher headline speeds and lower cost per byte. In the access market, connectivity and fixed and mobile networks have a life of their own; it’s driven by competition, this constant ramping up of speeds and capacity and so forth. It's not driven by end-user demand.

In the past, you could have made the case that demand trends will catch up in the end. But if, as we see, and in a complete reversal of the rather silly scissors diagrams we saw 15 or so years ago about the relationship between capacity and investment and ARPU in mobile networks, capacity is shooting ahead far faster than demand. And that probably generates tumbling ARPUs in the end.

In another podcast, we pointed out that 5G ironically has accelerated the decline in mobile revenue rather than added to it. It may be too many operators chasing too little demand and reducing prices in that way. Another point worth making here is that the additional capacity in fixed, once the passive parts of the network are in place, is very low, and it's effectively part of a maintenance cycle, so you will get large multiples of additional capacity generated at very low capex.

Future capex forecasts

Caroline Gabriel

So in all this context, can you describe what you're forecasting as a result of capex in mobile and fixed and why?

Rupert Wood

Yes. Separating it to mobile and fixed isn't the best way to do this. I think it's better to look at it in terms of infrastructure, so passive infrastructure, such as towers and fibre, and then technology - hardware, software and so forth. These have different costs, and critically, for any forecast, they have different useful asset lives and, hence, replacement cycles.

We think infrastructure, in the broadest sense, is about 60% of the total. It can vary between countries, but it's about 60% and an even higher proportion if you look at fixed-line telecoms. I think there are natural limits to what is required both on the fixed and mobile side, so there will be less overbuild in the long run. We can see that in the amount of overbuilding in fixed networks declining, and by replacing build. There's slackening demand for new wireless infrastructure at the same time, so its build-to-suit pipelines are narrowing over time.

We might also see, on the active side, rather than the infrastructure side, longer replacement cycles for actives based on slackening demand for actives and spectrum. I think we've already seen, in many cases, a rethink of 5G standalone by some operators, a slower impact of Open RAN and a slower transition to mobile network cloudification. I think there's little appetite for a major RAN refresh in 6G, and therefore, we don't expect a cyclical rebounding capex. Just because it's been cyclical over 2G, 3G, 4G and 5G, it doesn't mean it'll be cyclical again in 6G. I think the industry doesn't have the cash to do that.

Capex in these areas will exist, but it'll be more selective and more tactical than it was before. As a corollary to that, it's difficult to see spectrum remaining the kind of cash cow it is for governments now; I think we're moving away from that.

Caroline Gabriel

So how low do you think capex intensity will go?

Rupert Wood

Well, for operators, operators are a kind of subset of the industry in some ways because there are pure infrastructure plays as well, we think globally, it'll fall from about 19–20% to around 11 or 12% by 2030. That's a big fall. This doesn't include investing outside their traditional perimeter, so, they might invest in all kinds of other stuff.

If operators want to become essentially ServeCos and TechCos and leave their more physical asset base behind, then I think their investment will shift from capex to acquisition. It won't be the traditional capex that we see from integrated operators. We see current industry averages for operators that have divested their physical assets to a large extent at around 10 or 11% or even lower than that.

We think heavy wireline capex will be out of the system in most countries by 2030 and in many countries much earlier. And even if operators remain mobile operators in the traditional sense of the word, for mobile-centric operators at the moment, mobile capex tends to hover around about 12–13%, not those higher percentages you get from operators with fixed-line businesses.

Caroline Gabriel

I'd like to go back to something we touched on briefly: that much of this capex is not disappearing. It's simply getting transferred from traditional operators to other players, such as infrastructure companies, netcos, wireless infrastructure players, etc., and at the other end of the spectrum, even the cloud players.

Rupert Wood

Yes. Some of it is, but while delayering can increase willingness to invest at the margins and maybe improve some marginal business cases because you're collecting rent from more than one player, ideally, delayered entities still have to play within the demand constraints of the retail market. It doesn't magically create new demand. As for that shift to the cloud, well, cloud players have obvious capex economies of scale, and telcos can benefit from that in a way. But the kind of capex that is being shifted from in-house to the cloud has been nowhere near as large as infrastructure capex; it’s a different scale.

Caroline Gabriel

This rapid decline sounds like it will have different impacts on the various components of the industry, but in the end, isn't this all a blessing for operators?

Rupert Wood

Yes. Absolutely. It produces better cash flow coupled with hard-won opex efficiencies, so it's good. But it's really important that improved cash flow doesn't simply get invested in lower prices. A hyper-competitive market combined with lower costs tends to lead to deflation, and that's a big problem, so competition where competition is still strong. But I think having said that, operators will still have more of a choice, and let's hope that whatever benefit they derive won't get squandered on investment in margin.

Strategies for operators moving forward

Caroline Gabriel

What else could they do?

Rupert Wood

Well, some of the interesting options are around scale. One of the things I've suggested elsewhere is that there is that potential to buy back stakes carved out. But to do so perhaps in a cleverer way to spread investment across geographies with equity stakes across what are often geographical patchworks. Operators are buying back into the areas that are now owned in fixed networks by infrastructure funds largely.

For network businesses, there is a possibility we can invest in adjacent network infrastructure businesses. The obvious ones are data centres, edge and computility; those are ones that are widely discussed. There are other network businesses as well, such as EV charging, local renewable power generation, distribution and so forth. Those are all potentially adjacent businesses that could be invested in.

Then maybe there's a possibility to start thinking about reversing past trends; the past trends seem to have been for operators to retreat into the markets they're strongest in and not consolidate transnationally. So reverse that and consolidate transnationally but at a service or a B2B TechCo level, acquire selected businesses or acquire selected skillsets, investing in people skills which are tech skills. Those are potentially different ways in which they could go about doing things.

Challenges ahead for vendors

Caroline Gabriel

Looking at the industry more broadly, it sounds like it's going to be very tough for vendors.

Rupert Wood

Yes, without a doubt. This is what I've been trying to argue in this report, that this feels like the end of an era. An era of constant expansions of the productive forces that have reliably delivered faster, higher capacity networks and created traditional productivity gains, a constant lower cost per byte. I think that the era of constant productive forces is drawing rapidly to a close. There will still be localised opportunities for some of this equipment, but elsewhere, vendors will have to refocus on smarter, greener, perhaps more durable network technology. But perhaps in the end, and I'm consciously paraphrasing Chinese industrial digital strategy here, they will have to harness new quality productive forces.

Caroline Gabriel

On that note, sadly, we need to wrap up a fascinating discussion. Thank you so much, Rupert, for all your thoughts. There is more content on this topic on the Analysys Mason website. To automatically receive future episodes, please subscribe to the Analysys Mason podcast.