Navigating data traffic: technology upgrades, spectrum policy and operator strategies

Janette Stewart

Welcome to the Analysys Mason podcast. My name is Janette Stewart. I'm a Partner in Analysys Mason's consulting team in the UK, where I focus on regulation and policy, wireless networks and spectrum.

Caroline Gabriel

And I'm Caroline Gabriel. I'm a Partner in Analysys Mason's research team, also in the UK, and I lead our research on mobile and fixed networks and telecom cloud infrastructure.

Janette Stewart

Today, we will discuss some of the key themes in the wireless market we picked up from discussions at a recent conference we attended in Brussels. The conference was Forum Europe's 5G conference.

Some of the key themes included priorities for European policy and for the new commission on 5G, challenges in the telco market such as investment and moving to 5G standalone, 5G business models and evolution to 6G.

Two trends we were particularly interested in, which we wanted to talk about today, are wireless data traffic growth and the diverging views on that, as well as the implications of these diverging views on the spectrum needed for 5G and beyond. We'll also touch on spectrum for other wireless technologies that are key to providing short-range connectivity in the European market, such as Wi-Fi.

Caroline, do you want to kick off the discussion on the topic of wireless traffic patterns?

Wireless traffic patterns

Caroline Gabriel

Yes, sure. Of course traditionally, policy and operator strategies for spectrum have been heavily shaped by the assumption that mobile traffic would continue to grow globally at the kind of rates we've seen in the past, around 30% a year. However, as the conference discussions highlighted, there's now little consensus on future mobile traffic patterns. There are diverging views ranging from near-zero traffic growth in the next few years to significant new growth that is often driven by AI applications. Those might potentially also intensify a pattern of growth in traffic hotspots rather than network-wide increases. So, there are differences in the pattern of traffic as well as just the volume.

At Analysys Mason, we produce annual forecasts for traffic over fixed and mobile networks, and we relate the latter to spectrum and capacity requirements. We've seen that although 5G does drive an increase in the volume of cellular data traffic, the growth in that traffic has slowed since the introduction of 5G at the end of the last decade. We believe it will continue to do so, via a new set of applications or behaviours that are currently not really visible.

The cellular data traffic growth rate has fallen in every major regional block and the world as a whole in every year since 2020, except for North America, which saw increased growth rates from 2021 to 2022, mainly because of the launch and adoption of 5G fixed-wireless services rather than mobile usage.

The switch from LTE to 5G mobile does increase data traffic, but not by historic levels. Growth levels fell below 20% in 2024. And in some markets, we're even starting to predict absolute traffic decline, especially where there's heavy fibre and Wi-Fi expansion, such as in Mexico, or particularly high levels of mobile traffic that are always carried over Wi-Fi.

Of course, there's always the potential for a new type of application or user experience that will reverse the stagnation of cellular traffic growth. Many people – and there was a lot of discussion of this in Brussels – believe that these applications will be driven by the increasing accessibility of advanced AI on mobile devices. However, I think this can be overplayed. If such applications achieve significant adoption, it remains questionable whether they'll impact cellular traffic sufficiently to return to growth of 30% or so. To do so, they'd have to generate additional traffic on the cellular network specifically, and this could be limited by usage being mainly over fibre and Wi-Fi, as we see with most heavy-duty mobile apps currently.

We do expect smartphone-based AI apps to have some upward effect on traffic. However, some of this will be as users move from other activities, such as conventional video. After all, there's a limit to the amount of hours anyone can spend on their phone. Also, uptake may be gradual. So in Europe, for instance, we're predicting a compound annual growth rate of about 5% a year to 2029 in cellular access network traffic attributable to AI. But this compares to almost 80% CAGR in data-centre traffic driven by AI, so it's relatively manageable.

Also, we would expect to see a great deal of AI processing for consumer apps being localised on the device, see some of the work Qualcomm is doing on this, with only limited subsets of data needing to be transmitted across the network.

We expect 5G and future 6G traffic to become increasingly localised in hotspots and even in personal cells of capacity that are created virtually around each user as they move about. That's likely to be a key concept of 6G and one that will profoundly affect spectrum usage, traffic patterns and network planning.

Spectrum assignment implications

Janette Stewart

Thanks very much, Caroline, that's really interesting. Shall I pick up and just talk about what that means for spectrum assignment?

As you've said, the traditional assumption on spectrum estimation is that mobile traffic will continue to grow. It's also important to say that demand for new mobile spectrum bands also tends to follow mobile technology generation cycles. So, for example, the 800MHz and the 2.6GHz spectrum bands were new bands introduced when mobile network operators rolled out the fourth generation. 700MHz and 3.5GHz were the new bands auctioned in Europe for mobile in the fifth generation. Each new generation of technology has introduced RAN changes, and we'll talk more about that in a moment.

A key aspect of this technology change that was affecting spectrum demand is how the channel bandwidth has changed as mobile technologies evolve. 4G, for example, started using paired spectrum channels of 2×10MHz or 2×20MHz. In 5G, the evolution has been towards unpaired channels and much wider bandwidth. So 5G's new radio RAN uses 100MHz channels.

Currently, the only band that can offer this 100MHz channel is the 3.5GHz band. Other mobile bands, especially below 1GHz in the 1800MHz and 2100MHz range, are based on the legacy arrangements of paired 10MHz or 20MHz blocks.

This demand for wider unpaired channels is one of the drivers of future spectrum demand for mobiles. So, for example, if 6G uses 200MHz channels, then there's a need to find spectrum to accommodate that configuration of spectrum.

Notwithstanding that, the traditional approach when a new mobile spectrum is needed in a market is for the regulator to identify if the relevant band is available, to consider the demand based on growth in wireless traffic, and then to award licences via a competitive process, often an auction.

Regulators usually set a minimum price for licences to be auctioned, and that's called a reserve price. That reserve price can be influenced by various approaches, either using a business plan approach to estimate the value of the spectrum being offered to different types of operators or using benchmarks from previous auctions of similar types of spectrum.

Advising on both of those approaches, the business planning approach and the benchmarking approach to spectrum valuation, are things I spend a lot of my time doing in Analysys Mason, either advising the regulator or advising mobile operators.

We have a spectrum auction tracker product in which we track auction prices by mobile band and by market worldwide. From the auctions in our tracker, we can derive a benchmark of auction prices, which is usually expressed in USD per MHz per pop or equivalent. We can compare this benchmark with the proposed level of reserve price set by the regulator. We normalise for bandwidth offered, population of the market, licence duration and any other considerations such as annual licence fees or coverage obligations attached to the licence.

What's going to be interesting moving forward is that the auction benchmarks to date reflect market conditions in which operators need more spectrum for capacity reasons because of growth in mobile traffic. The business planning approach to spectrum valuation also typically looks at how mobile traffic will grow. In a business planning approach, we're traditionally calculating network costs and revenues for an operator over the licence term in a market in which traffic is growing. It’s therefore going to be an interesting change in market circumstances if operators are less pressured by traffic growth. What does that mean? Will benchmarks from previous auctions still be relevant? We're already seeing, from benchmarks, a trend towards spectrum prices reducing, for example.

Having said that, the other key driver of mobile spectrum demand is the evolution of the underlying RAN, and this is still likely to be a key driver. Caroline, shall I hand over to you to pick up on that point?

Evolution of radio technology

Caroline Gabriel

Yes, sure. It's an important one. I think the evolution of radio technology often gets forgotten in all the focus on demand, usage and traffic. That technology change will be an important one in spectrum considerations.

Some of the technology changes relate to improved compression and efficient delivery of complex media. Many innovations are ongoing in the cloud and media industry that can reduce the burden of video and AI content on the network and spectrum. Others relate to new radio technologies, they are more directly in control of the telecom ecosystem. And some of that may emerge in 6G standards.

There's a high expectation that 6G will specify wider channels of perhaps 200MHz, which could enable advanced applications and could drive operators to want to acquire a new spectrum that could accommodate these channels. However, even if these specifications emerge, some operators will need to see a very clear justification to buy the new spectrum.

Not only will they need to see new applications that will require these wide channels and will operate in cellular rather than Wi-Fi, but they will also need to identify a way to monetise them, which has been challenging for most of the advanced 5G and even 4G consumer apps from the telecom operator's point of view.

Otherwise, operators may be cautious about new spectrum and more focused on improving the usage, efficiency and monetisation of the spectrum they already have. Repurposing of legacy bands and increased technical advances to support key spectral efficiency techniques would allow operators to make far more out of their existing capacity. Those techniques include increased levels of spectrum reuse and carrier aggregation, plus dynamic spectrum access and dynamic aggregation of different bands on a on-the-fly basis, depending on the requirements of a certain application.

All that can be underpinned by an AI-enhanced software-defined radio, which we believe will be an important aspect of 6G-era RAN advances, and the concept of a network that is AI-driven and intelligent on both sides of the air interface.

Traditionally, operators have often acquired spectrum they didn't urgently need to ensure that they didn't end up at a disadvantage to their competitors. But as they become increasingly capex-sensitive, the era of 'build and they will come' is ending. In some countries, we've even seen operators collectively refusing to bid for spectrum that they consider too expensive to be justified by the business opportunity at hand. We saw some of that behaviour in India, for instance.

So, some operators will stay aloof from future spectrum auctions if they can't understand the return on investment. Alternatively, some operators may see a genuine requirement for wide channels and new spectrum, but perhaps in localised areas to support certain applications, communities or traffic hotspots, which in turn may drive a shift away from national licensing as the norm.

Spectrum policy implications

Janette Stewart

Thanks very much, Caroline. Let's move on and wrap up on some of the specific implications of spectrum policy.

As you've said, requirements for wide channels to support certain applications may create demand either in the wide area or in localised areas. Typically, in the wide area, we would expect the devices to use public network. In the localised area, there are various wireless choices, including Wi-Fi.

The concept of personal area networks that you described is interesting for future applications such as AR glasses. This is an interesting development, which may drive demand for these wider channels, not just in a public mobile network, to be supported by technologies such as Wi-Fi 7.

This conflict between the wide area and the local area traffic demand is at the heart of a big debate taking place in Europe at the moment, which the 5G conference talked about last week in Brussels concerning the upper 6GHz band and its future use. We've written a couple of articles on the upper 6GHz topic; please do look on our website for these if you are interested.

What's becoming clear from the upper 6GHz debate is that regulators are increasingly looking to technological solutions to manage future spectrum demand and to ensure spectrum is being used and assigned most effectively. Various types of spectrum sharing have been investigated in this upper 6GHz discussion. Location sharing, such as using the spectrum outdoors in public networks and indoors for Wi-Fi is one idea that has been floated. That approach may not be ideal for future personal area applications such as AR glasses, as it's when people are outdoors and on the move wearing the glasses when the connectivity is really needed.

The other version of spectrum sharing is more like a detect-and-avoid approach where technologies can sense each other and avoid frequency clashes. Thinking specifically about using upper 6GHz deployment in a mobile network, operators are likely to deploy that spectrum only in high-capacity locations. The question then for mobile spectrum auction design is how to package those licences for an auction since, if the spectrum isn't being used nationally, then auctioning national licences is unlikely to be efficient.

Add to that the fact that the spectrum might be shared with other localised applications such as Wi-Fi, and this raises questions about the best auction formats. Given that the coverage commitments that have usually been attached to spectrum licences have usually been for nationwide roll-out or some variation of that, then in future, there will be further consideration about how to establish coverage commitments.

Those coverage commitments may be decoupled from the spectrum auction, perhaps set more holistically by government policy rather than being specific to individual spectrum band roll-out.

There's more content on this topic on the Analysys Mason website. To automatically receive future episodes, please subscribe to the Analysys Mason podcast.