Let’s get one thing straight. Until just a few years ago, the author of this blog post assumed that he would be long retired before we started speaking about 4G/LTE switch-off, but the topic has started to be considered. And we are increasingly asked the question of when we expect it to happen.
Furthermore, in a recent report ‘Monetisation of public 5G networks through IoT’ we had to consider the continuing availability of LTE networks as a significant factor in the adoption of 5G for IoT. This is particularly relevant because one of the main conclusions of the report was that migration to 5G is overwhelmingly for the purposes of future-proofing, i.e. LTE would be perfectly adequate at supporting an application’s requirements, but it would not be possible to rely on the availability of networks for the lifetime of the device.
The logic of refarming less spectrally efficient 4G to more spectrally efficient 5G, as well as rationalising the number of networks being operated, is inescapable. The question is when? In this blog post we offer a few thoughts on the likely timings of 4G switch-off.
The first thing to note is that there will be a substantial geographical variation. There certainly was with 2G/3G switch-off and there is no reason why it would be any different with 4G. Back in 2020 we published a report ‘A global overview of planned 2G and 3G switch-off’ which highlighted the activities of operators around the world in refarming. At that stage, Australia, New Zealand, Japan, Canada and the US had all more or less completed their 2G switch-off processes and were well under way with 3G. In the case of Japan, 2G had been off for almost a decade. Contrast with Europe where today 3G decommissioning has been fairly well progressed but 2G remains largely intact. If this approach plays out also in 4G, we can expect any process of refarming/switch-off to be stretched out over a period of 20 years or more between the vanguard (probably the US, Australia and Japan) and the laggards in Europe and elsewhere.
There will also be much variation between individual MNOs based in large part on what spectrum they have. For instance, those that are struggling competitively due to lack of appropriate spectrum may need to accelerate their refarming in order to keep up with those that have better spectrum resources.
We should also note that refarming isn’t all-or-nothing. We can expect, as we have seen with 2G/3G, that there is a gradual deprecation of the network as certain frequency bands are reallocated or sites taken offline. This generally has only a modest impact on the experience of the technology for most users. Eventually, however, the whole network will be shut down in a rather more binary way, thereby allowing the switching-off of core network infrastructure. As a process, I’m put in mind of an exchange from Ernest Hemingway’s The Sun Also Rises:
“How did you go bankrupt?” Bill asked.
“Two ways,” Mike said. “Gradually and then suddenly.”
Similarly there may be a choice to cease support for particular categories of LTE, downgrading the capabilities (and therefore the required capacity) of devices.
And there is a technical solution which also offers a half-way house. Through Dynamic Spectrum Sharing (DSS), the same RAN can be used to support both LTE and 5G. It is entirely possible to maintain support for 4G via what is ostensibly a 5G network, although it would currently require a 4G core. However, DSS is somewhat (perhaps 20-30%) less efficient for both 4G and 5G, which all but negates the benefit using it.
In this blog post we consider the likely timing for the first few markets, rather than the bulk. And we consider the timing of the point at which the 4G functionality is deprecated sufficiently to have a significant impact on coverage, rather than the point at which some refarming occurs (we’re already past that point in some markets) or at the other extreme that the technology becomes entirely unavailable. Furthermore, we would expect sales of 4G devices to decline quite rapidly in the 5 years running up to a final switch off.
Taking the macro view of mobile generations, a lifespan of 30 years might be reasonable, taking LTE out to about 2040. Generally speaking, when considering the timing of the switch-off of 2G/3G by companies in the vanguard of that process, 2G lasted for around 30-35 years and 3G for around 20-25 years. There’s some complexity here. In Japan, 2G lasted only a little over 20 years, but that wasn’t using the GSM standard, so couldn’t tap into the global scale of that ecosystem, hence a greater desire to move to more universally adopted technologies. The first LTE networks were switched on in 2010. We’d expect lifespan of LTE to be more like 2G than 3G because of how successful the technology has been. History points to the late 2030s/2040.
But we can’t rely on history, so we should consider drivers and barriers.
The main driver for refarming 4G to 5G is to free up spectrum for a more spectrally efficient technology and to reduce the operational cost of running two sets of core network infrastructure. That being the case, the quantum of the efficiency and cost savings becomes very significant. It’s interesting to compare the relative benefit of 4G refarming to that of 2G/3G.
The spectral efficiency gains of refarming 2G/3G is significantly better than of refarming 4G. There are many vagaries about the specific impact, but going from 2G to 4G gives a greater than 10x efficiency gain, in terms of bits/Hz. From 4G to 5G is around 2-3 times more efficient. There’s also a question of revenue per Hz to be made too. If the only way to monetise the greater efficiency of 5G is by selling enormous Fixed Wireless Access data plans at knock-down prices one must ask the question of whether all that spectral efficiency is really being properly monetised.
The cost savings of knocking out the requirement to support LTE network infrastructure is, however, clear. Although, again, the benefit of removing 2G/3G would have been significantly higher, just because it relates to rationalising out two network technologies rather than one. We also note that there are very cheap ways to run virtual EPCs meaning that the core network savings of removing 4G are minimal.
The great migration away from 2G/3G has happened or is happening, and there are doubtless significant cost savings associated with that. What we are sceptical about is whether the further rationalisation by switching out 4G would result in anything like the benefits from 2G/3G. If it doesn’t why would we expect the process to be faster than for 2G/3G?
One of the big headaches with refarming spectrum is the installed base of devices making use of that legacy infrastructure. Specifically IoT devices that would need to be replaced.
The year 2022 was ‘peak 2G IoT’ globally, with about 320 million devices installed. In comparison, by 2032 we will hit peak 4G IoT, with 1.8 billion devices globally. Obviously, the dynamics will be rather different by country, but with a similar picture. 4G will be much more widely deployed when it comes to switch it off than 2G was. See our Forecast Highlights page for more details.
To take one specific example, consider the 100 million vehicles in the US that Transforma Insights estimates are connected using 4G. What would be the cost of replacing those devices? This only being a blog post rather than a full report I’ll satisfy myself with the answer “substantial” and leave it at that.
There’s a bigger volume of devices now on LTE networks than there were on 2G/3G when their switch-off was announced. So it’s a more painful exercise now than it has been in the past. Of course it’s debatable whether this is actually a driver for refarming more rapidly or more slowly. If you wait, the volumes will only continue to increase.
We should also note that there is some question of future-proofing. LTE-M and NB-IoT are now part of the 5G standard with the ability to be supported on 5G core networks. So, part of what was ostensibly 4G will be supported on 5G.
However, there is some further complexity. In many cases, notably in the US, the MNOs have not moved LTE-M traffic to the 5G core and will continue to support via the LTE EPC. Furthermore, LTE-M devices would need to explicitly support 5G core signalling, something which is not generally happening, for instance with Qualcomm’s 9205 chipset not supporting it. The result is that for some carriers, LTE-M remains, in effect, an LTE technology.
That being the case, we can envisage a scenario in some markets, notably the US, where the MNOs might only maintain LTE-M, supported on an LTE EPC, for a decade or more (two major US carriers have committed to support LTE-M on an LTE EPC until 2035) and then refarm the spectrum to the next generation of 5G (or even 6G) aimed at supporting low power connections; specifically the further revisions on 5G Redcap. But that is quite unlikely, given the widespread use of LTE-M as the low power technology of choice in the US and the relatively small amount of bandwidth required (1.4MHz).
In our recent report ‘Monetisation of public 5G networks through IoT’ we examined the price of 5G devices relative to 4G. Currently there is still a substantial price premium associated with 5G hardware compared to its LTE equivalent. For instance, currently RedCap competes with LTE Cat 4 (and eRedCap will compete with Cat 1 as the next wave). Today RedCap prices are 2-3 times that of LTE Cat 4. Prices will need to come down much further for there to be an appetite to switch. Realistically there needs to be no more than about a 20% premium for customers to be willing to migrate from LTE of their own volition. Forced migration due to network switch-off would of course accelerate the trend but will likely result in both disgruntled customers and slower adoption.
The timing for anything approaching price parity with LTE is completely unclear. With scale we do expect prices to come down significantly. And several MNOs have initiatives to drive greater scale. However, inevitably due to the additional complexity there will be some premium compared to the LTE equivalent. We should note that 5G devices will become cheaper if there is no need to support LTE fallback, as for instance is the case with RedCap devices today. By the mid-2030s however, we might see pricing sufficiently low to allow for migration to 5G.
We should also note that there will be a price premium for what would otherwise have been 4G-only customers. With the exception of Australia, Japan, the US and a couple of other countries 4G switch-off probably won’t happen before 2040. And at the same time 5G network coverage mostly won’t be on a par with 4G. As a result, any IoT deployment involving more than one or two specific countries will inevitably need to support LTE, as the most universally deployed technology globally. So anyone building IoT solutions would need either multi-mode 4G/5G or multiple SKUs.
In Europe, MNOs are only just making inroads into switching off 3G, a technology that is ostensibly not required for either coverage or capacity and is rarely relied on for IoT. 2G switch-off is on the horizon, but for most major European countries support for 2G is likely to continue into the 2030s. 4G switch-off is unlikely before the mid-2040s, we expect. Emerging markets in Africa, Asia and Latin America will likely follow Europe.
In contrast, in the US, the major MNOs are showing increasing concern over pressure on spectrum and there is likely to be a move to refarm LTE spectrum for 5G within the next decade. We can anticipate that a few other countries may take a similar approach: Australia, Japan and South Korea have all been in the vanguard of 2G and 3G network refarming and may take a similar approach to the US.
