5G / 6G network efficiency may grow faster than data demand. Overcapacity is a serious risk

Technological improvements in wireless networks are yielding faster growth in mobile data capacity using existing spectrum and sites, compared to top-level growth in demand and traffic volumes.

In other words, expected future data traffic should mostly be handled by more spectrally-efficient networks. This means that there is little or no need for extra exclusive-licensed, wide-area spectrum for 5G and 6G, given current demand patterns and reasonable forecasts.

Furthermore, future mobile data volumes may well trend towards low or zero growth, according to some more bearish forecasts. Recently reported traffic figures from Europe, China and elsewhere suggest many "mainstream" industry predictions are overoptimistic.

Indeed, there are even reasons to believe that mobile traffic could fall in absolute terms.

If these factors turn out to be correct, the mobile industry potentially faces the spectre of overcapacity, especially if additional spectrum is acquired along with rollout mandated by the licensing rules.

This is also one of the scenarios considered at my upcoming Unthinkable Lab workshop in London on April 8th, run with Andrew Collinson. The industry - and its investors and regulators - need to consider many risks and opportunities, to shake a "stuck" telecoms sector out of its torpor, and to mitigate the harm of external factors such as geopolitical volatility.

Demand growth and forecasts

We've gone well past the point at which anyone can claim that mobile data growth is "exponential" and remain credible. It clearly is not still on a hockey-stick curve. Neither is it "exploding", "inexorable" or any number of similar hyperbolic adjectives.

In most parts of the world, overall mobile data traffic growth is now at 10-20% a year including fixed wireless access. This means it is in the range 0-10% growth for mobile broadband / smartphone data, if you strip out FWA, which creates maybe 20-30x the traffic per subscriber as MBB for the same revenue and skews the aggregate data.

I've included Ericsson's most recent predictions below - which can be taken to be near the upper bounds of a reasonable forecast range.

Ofcom's latest report for Q3'2024 had overall mobile data up 7.7% year on year, including FWA. In France, ARCEP put Q3 growth at 12.3%, down from 18.2% a year earlier. Looking at Ireland's regulator ComReg, it looks like it had around 3% growth in mobile data in 2024, although it's a little hard to be certain as the quarterly reports seem to have had a definition change midyear. It certainly doesn't align with its consultants' central prediction (Frontier Economics) of 16% CAGR 2023-28, which looks significantly overcooked. Worse, that was fed into its spectrum policy.

As I noted in a recent post, China Mobile's latest figures showed that it had completely flat data use per handsets in 2024 vs. 2023, and the last quarter actually showed a small year-on-year decline in absolute data traffic. Similar trends have been observed in markets such as Finland.

Notably, this is despite China having wide availabity of 5G standalone. There has been no uptick from slicing, XR and all the other fancier features often touted as driving traffic growth. Even with lots of IoT, V2X, surveillance cameras and a fair amount of FWA, the country's MIIT reported total overall mobile data traffic growing 11.6% in 2024, a figure that has fallen from 14.6% in the first half of 2023.

Some observers such as William Webb think everywhere will trend towards zero growth in data traffic. Others note the differences in traffic per subscriber between national markets and suggest that shows scope for "laggard" countries to level-up their consumption to match their peers. That seems unlikely to me, and probably reflects other differences like demographics, prevalence of fixed/Wi-Fi connectivity or even different qualities of indoor coverage.

For now, let's take 0-10%/year as a reasonable range for handset data growth, and 10-20% per year for overall mobile data. (For reference, that's broadly aligned with fixed broadband growth). I'll come back later to possible reasons we might see absolute falls in data traffic.

Spectrum and network efficiency improvements

But as well as traffic trends and forecasts, there's another dimension here: efficiency of networks at squeezing in more data throughput, for a given set of resources. There's a huge emphasis on this, especially given the cost of adding capacity and spectrum. It's also a prime target for wireless-sector AI specialists and their suppliers.

Various technological developments for 5G and emerging 6G prototype networks are increasing the data-carrying capacity of today’s networks. Many individual advances offer promised improvements of 20%, 30% or even more in spectral efficiency or other gains – and can be combined together to give an ongoing pathway to more effective usage of both networks and spectrum resources.

A key trend in 5G networks is what is called “massive MIMO”, or multiple-in, multiple-out (MIMO) radios. These have arrays of small antennas that can “beamform” and increase overall capacity of a given site significantly. Rather than send signals uniformly across a wide arc, these use clever radio techniques to combine many transmissions into a more directed beam. It’s not really the same as a laser compared to a lightbulb, but that is a helpful visualisation.

Commonly, leading MNOs deploy their midband 5G networks with antennas that use 32 or 64 elements for transmitting and receiving signals (32T32R or 64T64R). Various studies have suggested that the 64-way systems are perhaps 20-30% more efficient than 32-way, which themselves may have already given a 3x improvements on earlier 4G 4T4R or 8T8R radios.

Early research and trials have started into 128T128R and higher MIMO orders, as an improvement to this. (e.g. Samsung and Nokia) which could yield additional gains in future.

Other recent developments and examples enhancing spectrum efficiency include:

• Remote, automated tilting of antennas, optimised by AI techniques have delivered 25% spectrum efficiency gains for Rakuten Rakuten Symphony in Japan, allowing it to service high-consuming end users, with limited spectrum resources.

• AI-enabled “channel estimation” allows 5G radios to better understand how signals propagate in the real world, taking into account factors such as user mobility and interference. A recent announcement by Mavenir and Aira Technologies suggests 35% potential improvements.

• Cohere Technologies has been pushing its "universal spectrum multiplier" technology for some time. It claims a potential gain of up to 50% from software and AI techniques.

• Globalstar's XCOM Labs has a multipoint radio system that claims 4x gains, although today mostly for indoor / private 5G use.

• Ericsson has long been an advocate of sharing spectrum between an operator's 4G and 5G networks, allowing maximal use of existing resources. A similar technique can be expected for 6G

• 6G is expected to include efforts to commercialise Distributed MIMO, where the antenna arrays are split across separate physical units and synchronised to give further improvements over more monolthic designs. Samsung and KDDI announced an AI + dMIMO project recently. This ties in with broader #6G visions arond "cell free" or cooperative architectures.

• Digital twins of networks and radio propagation will improve network design, operation and understanding of interference, MIMO and other variables. This is a big part of NVIDIA AI-RAN concepts - and something that makes much more sense than using the RAN for GPU-aaS models for customer workloads.

• There is ongoing interest in RIS (Reconfigurable Intelligent Surfaces) that could reflect, shape and direct propagation of signals. While it doesn't generate extra capacity, it ensures that existing radio is distributed to end-users more effectively. While there are numerous practical constraints, it is another example of the ability to squeeze more out of a given spectrum band.

All of these - and others - point to a thriving domain of both existing products and new research that is yielding improvements to network capacity.

While not all will succeed, and some may just be applicable in certain places or user scenarious, it seems quite reasonable to suggest there is a roadmap for perhaps 15-20% a year - and maybe a bit more - for overall #5G / #6G efficiency gain, probably into the 2030s.

Not all those technologies may actually get deployed everywhere - but that's just because the industry often prefers the simpler method of "just throw more spectrum at it". WIth growing competition and other demands for frequency resources, that's not a sustainable solution.

Other shifts affecting supply and demand for 5G / 6G

There are assorted other technology and behavioural shifts at play here as well, which could also change the supply / demand balance for mobile:

• More auto-connect #WiFi, for instance using Wireless Broadband Alliance (WBA) #OpenRoaming as well as #PassPoint. This could further shift the existing balance of cellular/WiFi usage away from macro RAN consumption, especially for indoor use.

• More dedicated indoor cellular systems, using DAS, small-cells, #neutralhost architectures and so on. While this is inherently actually a really good thing for building owners and users, and could even drive up total traffic volumes locally, it will also offload traffic from the outdoor macro networks.

• Emphasis by some operators to persuade regulatory change on data compression or "data austerity". Some MNOs such as Vodafone are advocating for regulators to force content companies to use more aggressive compression, or let the networks "optimise" traffic on their own. They are still looking through the lens of a "capacity crunch" or fallacies about "exponential growth" rather than the reality of flattening demand. They often cite the unrealistic 20-30% industry forecast fantasies, rather than their own actual (and likely more modest) demand growth and internal predictions.

• More refarming of operator's old 2G, 3G and soon 4G spectrum bands for 5G and 6G, reducing the need for new spectrum to handle any increased usage.

• FWA is becoming the main source of traffic growth. But it tends to be limited to specific places with little FTTH / cable availability or competition. It can also be served by new extended-range mmWave spectrum, or local or shared spectrum options such as CBRS, 6GHz+AFC or even unlicensed bands.

• Spectrum-sharing or pooling arrangements tend to yield higher overall efficiency

• Operators are looking to consolidate. If the VF/3UK example is one to be followed, that could mean agreement with competition authorities to add more capacity, via promises to invest more

There is also another unknown - the potential for AI-driven compression or changes in network usage patterns. While some observers contend that GenAI could lead to higher mobile data usage, I think there are some reasons to think the opposite might occur. Without going into full details, consider:

• Use of #AgenticAI reducing the amount of traffic to and from the end-user, by automating tasks "inside the cloud"

• Semantic compression could yield up to a 10x improvement over traditional encoding schemes, by just sending descriptions and meanings of images or video, rather than pixel-by-pixel raw data. A basic step could be reducing pixel counts or frame-rate in unimportant parts of a scene for a video or game.

• On-device AI could create (or encode) content or information locally, rather than needing access to the cloud for everything.

5G Overcapacity: An Unthinkable scenario?

The bottom line is:

• Data traffic demand is slowing, perhaps to zero, or at least a very low % annual rate

• Data connectivity supply is increasing, given new sites & also network efficiency gains

• Demand might even fall in absolute terms, at least in some places, because of AI, Wi-Fi use etc

• Supply might expand still further, if MNOs get extra spectrum & are then mandated to deploy infrastructure as part of license terms.

This is no longer a fringe scenario. There is a genuine risk of 5G / 6G overcapacity, which would have all sort of knock-on effects. I'm just musing here, as obviously there are many factors involved, but consider these to be possible:

• Likely price wars between MNOs

• Reluctance to invest in yet more capacity, but instead in efficiency to match rivals

• Operators handing back spectrum, which could impact pricing or even invite new entrants

• Creative ways to "soak up" excess capacity, such as cheap FWA, mobile-connected cameras, more MVNOs and wholesale services etc

• Large impacts on vendors

• Potential rethink or delay of 6G

• Location-specific (or time-specific) increments in capacity, not nationwide

As a sidenote, we already see examples of overcapacity in fixed networks - some places have multiple overbuilt FTTH networks in the same streets, each with low uptake that doesn't yield profitability. And where people do have gigabit or faster connections, most of the capacity remains unused and effectively unvalued by customers. This is leading some operators to try to put "artificial scarcity" at key chokepoints, such as interconnection or attempts to change neutrality rules. Rather than look at ways to drive demand, they are trying to constrain supply.

Policy implications

Probably the most important consideration here is "don't add fuel to the fire". Policymakers should certainly incentivise greater efficiency in use of network infrastructure, spectrum and energy - but should also think through the implications.

In particular, additional releases of spectrum for public mobile use by MNOs should be carefully scrutinised, given the many other demands. There is certainly no rush.

All regulators, policymakers (and investors) should revisit their demand forecasts for spectrum & network expansion, as many are based on outdated or unrealistic mobile data forecasts. Any document talking of "exponential" growth should be considered useless.

Clearing entire bands for auctions was already looking anachronistic. Now it looks harmful to both existing users AND the mobile industry. Given the difficulty of moving incumbents and the growing clamour for more spectrum from other groups such as satellite, military and unlicensed users, policymakers should instead be looking at – and potentially incentivising – a pipeline of spectrum efficiency improvements for wireless networks.

This would mean:

• No need for expensive or hasty clearance of spectrum with hard-to-move incumbents, such as defense systems

• Allowing time to see how 5G usage actually develops in the real world, rather than relying on unclear or questionable predictions. Gather more evidence on usage patterns, locations, competition and other metrics – and align spectrum with evidence rather than assertions.

• Less risk of network overcapacity, if data demand actually flattens or falls

• More deployment of Open RAN networks, which can support better “programmability” functions, such as using innovative spectrum and energy efficiency software.

• FWA can be delivered using various spectrum bands, and its use skews the overall traffic statistic. It does not need nationwide uniformity either.

A further set of policy issues here concern energy efficiency, AI regulation, cybersecurity and national security and resilience. Those are topics for other articles to address in depth.

Conclusions

Overcapacity is a genuine and realistic scenario for mobile networks. Maybe not everywhere, and maybe not imminently, but this is an Unthinkable Scenario, not an implausible one.

At the very least, it seems entirely possible that network / spectrum efficiency will keep pace with mobile data demand.

If traffic growth falls to 10% or below, as seems likely, then networks will rarely be running "hot" and investment should shift to coverage gaps, rather than capacity.

And under some possible futures "data deflation" becomes a real risk - demand falling, while supply increases. That would bring some very significant impacts and second-order consequences.

The recent announcements about both data growth and promising technologies for spectrum / network efficiency mean this is no longer an abstract, hypothetical scenario. It is real, and may get worse.

Operators, investors, policymakers and vendors need to start addressing this immediately.

If you want to discuss this in person - and many other interesting scenarios and opportunities - it will be one of the themes at the first Unthinkable Lab on April 8th in London. Full details (expand the agenda etc) and registration here. Otherwise, please feel free to discuss in the comments, or reach out to me directly for private advisory input.