In the Internet of Things we talk quite a lot about IT/OT convergence, the idea of taking a particular vertical or sub-vertical’s esoteric operational technology (think, for instance, of industrial control systems) and rendering them into the IT environment, i.e. available for developers to integrate with ERP, CRM, workflows and so forth. Last week I was in Las Vegas for the US variant of Mobile World Congress and many of the conversations crystalised a theme I’ve been thinking about for a while, the need for telecoms to go through its own IT/OT convergence. We don’t tend to think of telecoms networks as operational technology but I’d argue that many of the same ideas and limitations apply. It has its own language that is not easily penetrable for the IT/developer community and it’s not necessarily always easy to abstract into the IT domain.
What we’ve been seeing for a few years is a gradual move towards IT/telco convergence. The most prominent areas in which that manifests itself is in the new API initiative that was widely discussed. 5G also does a similar thing with Network Exposure Functions. And we’ve also seen some over-the-top efforts, for instance with Soracom’s Polaris and Dipper features, or 1NCE’s certification of developers. You can find out more about those and other forays from IoT connectivity providers into the IT/developer domain in the Communications Service Provider (CSP) IoT Peer Benchmarking 2024 report (login required).
In this blog post I explore a range of the things that I saw and discussed at MWC Las Vegas, some of which are completely unrelated to that overarching theme.
First thing on my agenda at MWCLV was the Open 5G IoT Forum. It wouldn’t quite be right of me to share my own presentation as being part of the key themes, for reasons of circularity and humility. Anyone who was there who was interested in the Transition Topics I talked about, you can check out the recording of a presentation I gave at MWC Barcelona where I discussed it, here. For those interested in the 24 key trends in IoT connectivity coming from our CSP Benchmarking report, here’s a blog post I wrote about some of the themes: Good ideas and bad ideas for Communications Service Providers in IoT.
But there were some interesting discussions in that room on that first morning that set the scene for some of the key themes. Cameron Coursey of AT&T shared particularly a couple of use cases which set the scene for me for a couple of trends that are out there at the moment. He was talking about – at the extreme ends of the IoT sophistication spectrum – the AT&T Smart Label and Video Intelligence. For former we have recently published some forecasts on, as discussed in a blog post that came out only last week: 'Disposable Devices: LPWA driving the booming single-use trackers market'. The latter is an area that we have delved into in great detail, including in this report: ‘Video analysis has significant potential with business benefits spanning across multiple industries’. So far so good in terms of covering the types of technology that we at Transforma Insights have well covered. But, as I say, the interesting thing is more that they also impinge on some bigger trends, both of which are considered below.
First AI. It will be unsurprising to anyone involved in telecoms that a lot of the focus of discussion at MWC LV was on AI. I should give a (second) reference to Soracom for a neat AI no-code platform for IoT applications, Soracom FLUX. When the GSMA folks were asking for a favourite demo/product over the week for the analyst wrap up session on Thursday, this was the one that sprung to mind. And there were plenty of other discussions of AI, whether it be for customer acquisition or network management. Or there was a very cool example of data normalisation and analysis from NTT in the form of Ultralight Edge AI, which localises AI models onto the local device which can pull in data from diverse sets of supplier equipment (e.g. industrial equipment, building management etc.), puts it into time series, contextualises it and runs AI models on it. It supports over 300 equipment vendors. Back in June we published a report on how MNOs/MVNOs were using AI for their operations, which I thought was quite interesting. In summary: mostly for the same stuff as every other vertical. See ‘How are MNOs and MVNOs harnessing AI for their IoT operations?’ for more details (log in required).
But it’s not really that form of AI that I was really looking for. What I’m increasingly on the look out for is what role the telco can play in supporting the delivery of AI. It was something that I raised in my thoughts on T-Mobile’s Capital Markets Day. Specifically it’s the role of the telco as an orchestrator of AI. AI processing and data storage can happen at any one of perhaps 5 or 6 levels in the network, from the actual end point through to a centralised cloud (or on-prem) function. Any deployment has to act on principles of subsidiarity, i.e. borrowing from socio-political thought the idea that decisions should be made at the lowest level possible. In the context of AI maybe it’s more that they be made at the most appropriate level. But the idea still holds. Orchestrating the level at which data is stored and processed and the AI instance sits is going to be somewhat complex and I think there’s a role there for the telco. But I’m yet to hear anyone really understand that point. Every discussion descends into consideration of using AI for network management and similar functions. That’s important but it’s not a differentiator. Everyone will do it and sweat their assets better and some (maybe T-Mobile) will get a year or two of competitive advantage. The orchestrator role is more fundamental to how AI is delivered. Providing the ability to deliver AI more cost-effectively and/or with lower latency is a valuable business service, and one that is potentially chargeable at something other than cost-plus. Value-based charging anyone? (More on this in the summary, below).
If you’re interested in this convergence of AI and IoT, including both the internal use cases noted above, and the orchestration opportunity, check out the blog post ‘AI and IoT: a symbiotic relationship critical to meet enterprise needs’.
And now onto something rather more mundane and . One of my topics for discussion in my scene setting at the IoT session related to apathy towards NB-IoT and (to a lesser extent) LTE-M. This is interesting in the context of the discussion around low power trackers, which traditionally tend to favour those technologies.
It turns out that my hypothesis of apathy is perhaps even putting it too gently. Rumour has it that there’s likely to be some NB-IoT sunsetting quite soon. I had quite a few conversations about this. It seems like NTT Docomo was ahead of its time when it mothballed NB-IoT in 2020. To be honest I don’t think it’s as simple as the lights going out for NB. It’ll be a country-by-country decision. Most NB seems to be deployed for single country deployments anyway (smart metering for instance). Any operator supporting one of those, or for instance the break down beacons in Spain, will continue to run the network.
But at the same time I don’t think 5G RedCap is anywhere near ready to take the strain, for reasons I’ve laid out before, most recently here: Good ideas and bad ideas for Communications Service Providers in IoT. Maybe with eRedCap, more likely with a further iteration, it will be but today it’s not really optimised for price, power or functionality in a way that makes it optimal for the major IoT use cases.
And while we’re on the subject of sunsetting, the question of 4G switch-off came up too. And specifically timescales. I don’t think anyone would expect LTE to be around in perpetuity, but I also think that some of the predictions of accelerated switch-off are a little wide of the mark. 2G generally lasted 30-35 years, 3G 20-25 years. Longer in many countries. I’d expect LTE to be more like 2G rather than 3G, because it’s good (like 2G) rather than bad (like 3G)…if you’ll excuse the exceptionally crude analysis. So just on that basis late 2030s would seem reasonable. And then it’s likely to be the usual suspects in US, Canada, Australia, Japan and South Korea making the switch rather than in Europe. And these things are never 100% switch-off on a specific day, it’s gradual diminution of the network. And there’s numerous other reasons: devices costs are still high, for instance, and the number of IoT connections running LTE is now bigger than the 2G/3G base was so the switching costs will have grown. Also, and this could do with me actually working out the exact figures, but rationalising out 2G and 3G networks (i.e. from 4 network technologies to 2, typically) with the removal of multiple infrastructures, offers better cost savings than the equivalent for 4G (i.e. from 2 to 1), And, the spectral efficiency gains of refarming 2G/3G is significantly better than of refarming 4G. And I’d argue that it’s diminishing returns. The value of the additional services you can deliver over 5G is not radically better than LTE, at least not proven. And, as we commented in our report ‘LTE Cat 1 bis is a real contender in cellular-based IoT, at least in the medium term’, “two major US carriers have committed to support LTE-M on an LTE EPC until 2035”. And there’s a bunch of other reasons, meaning that some of the more aggressive predictions are I think not right. I realise I’m putting my head above the parapet a bit here, but late 2030s seems like the most realistic timing for LTE switch-off in the first markets.
From a potential source of doom and gloom to what I see as being one of the bright spots. I also had the opportunity to sit down with a range of companies to talk SGP.32 remote SIM provisioning. And I was very complimentary about it in both of my sessions. It strips out a lot of the complexity of handling RSP and that is to be lauded.
The interesting thing is the impact that it might have on the structure of the market. What we’re increasingly seeing is companies positioning themselves as eSIM connectivity platforms (the terminology needs some work) managing connectivity localisation onto local networks on behalf of the enterprise, with the enterprise able to avail themselves of bring-your-own-connectivity (BYOC), although most of them not doing so initially. I had some very interesting chats with Simplex Wireless, Teal Communications and others. These companies are now positioning themselves as something of an interface between the enterprise and the whole connectivity ecosystem. But it’s still a managed service rather than a free-for-all, which is rather what the MNOs were worried about.
And while you’re here, I should give a plug to our webinar on 16th October (also to be available on-demand) with Eseye ‘What enterprises need to know about using SGP.32 ‘IoT’ remote SIM provisioning’ where we’ll provide a reality check on where we are with SGP.32 and the impact on IoT connectivity. Register HERE.
But, if there was one over-riding theme of the event it’s that idea of telecom’s own IT/OT convergence, rendering the telecoms space into the IT domain and as a result into the enterprise’s domain. What we’re perhaps starting to see emerging is an approach of ‘enterprise-down’ rather than ‘network-up’. At a very basic level it’s about thinking from the perspective of the customer rather than a narrow product view. For instance, I had a good conversation with Cisco about a range of things, but most relevant was the fact that it is now getting the hang of cross-selling (and co-selling with its Control Center operators) CC-based connectivity with its industrial routers business.
Simetric is a good example of this trend. The focus is on offering a set of functionality that meets the requirements of the end customer rather than acting as another telecoms layer. It provides abstraction of connectivity and device management, with a set of value-added services layered on, including integration with workflow and operations management. The trigger for the availability of this kind of functionality could perhaps be thought of as being the abstractability of the various functions. Can you pull the data into a single management platform? And will we soon be able to do that kind of thing with network functions too? Perhaps.
The quintessential example of all of this is probably all the talk about network APIs, and their opening up to developers. There was a lot of chat about this but it’s still at a very early stage, as it is with 5G network functions. We’re at the start of a journey, but a positive one. And hopefully one that sees a pivot by telecoms from scrapping around for a share of a diminishing market to looking to deliver something more appropriately tailored for enterprise customers.
In the telco stack everyone is nickel-and-diming everyone else (see monthly recurring charges, network attach fees, general antagonism from MNOs to MVNOs, lack of ability on the part of MNOs to collaborate, and so forth), and selling to the customer on pretty much a transactional basis of lowest cost per-MB. This is the reason why so many IoT connectivity providers are moving into additional value-added services and consulting; they have to. Perhaps the new environment of native IT-centricity (as opposed to the over-the-top rendering and abstraction) offers a new model.
