Phil Skipper, Head of Business Development at Vodafone IoT, was first involved in IoT back in the heady days when it received the kinds of forecasts that AI does today. Perhaps it’s a sign of things to come that he avoids talking too much about the technology. That’s not what his customers are buying except incidentally.
“Customers buy something which is very different, which is what we call the confidence to connect,” he told TelcoForge.
“We offer a managed IoT service, so all of that complexity about what technology is used where happens below the waterline for the customer.”
The customer-facing message is stripped to a single sentence – “we connect anything anywhere on the most suitable technology” – and the underlying mix of 2G, 5G standalone, NB-IoT, satellite, and anything coming next is entirely concealed “so that our customer doesn’t need to become an expert in our business to run their business.”
This is quite a contrast to some early IoT pioneers, such as Verizon who offered home security and other IoT services in the very early days and didn’t see it take off. Skipper described this as “organisations trying to reach up too high into areas where they, by definition, have a lack of knowledge.”
While Vodafone IoT is explicitly not following this course, the inverse is also true: they work hard to keep customers from reaching down into Vodafone’s own area of expertise. The technology isn’t simplified for the customer. It’s hidden from them.
This holds across customer size. A multinational car manufacturer and an SME with five security cameras are, on Skipper’s account, buying the same underlying thing (“but they may want to buy it slightly differently”) and what both actually want isn’t IoT at all:
“What they actually want is an incredibly reliable connection that does what it says on the card every single day of the week… we call it IoT, but customers buy something which is very different.”
For the SME specifically, the pitch is closer to a consumer transaction than a technical procurement: “they come to us and they say, I want five SIM cards and five security cameras, we can go, okay, here it is, it’s in the post, done.”
Skipper doubles down on the point “If you imagine you take your phone, you pop down to a shop, they give you a SIM card, you plug it in and you can make phone calls. There’s nothing different with IoT.” The sell is essentially a phone contract, not a system integration.
The same pattern holds for the two areas where Vodafone IoT is investing hardest.
AI, to Skipper, is something Vodafone IoT will consume to make its own invisible machinery run better – “We are going to be a massive consumer of AI ourselves because it enables us to do really clever things about managing every connection” – not something it exposes to the customer as a feature.
He rules out the application layer explicitly: “Do I think I’ll ever become the producer of incredibly complex AI driven security cameras? No. Do I think I’ll be providing connectivity that’s designed and optimized for AI driven AI security cameras? Yes, I do.”
Security works the same way. It isn’t sold, or even really discussed with the customer as a decision point – it’s bundled invisibly into the reliability promise. “If you take our managed service, you get security because we don’t want to be operating less-secure systems. So that comes for free.”
Even the forward-looking post-quantum cryptography work gets folded into the same “you don’t need to think about this” positioning: the quantum-safe VPN is already shipping in Vodafone’s hardware, well ahead of any customer asking for it. The technology is never presented as a product feature, just a promise being kept.
Model Player
The mechanism that makes this possible, in Skipper’s account, is refusing to sell anything else. Vodafone IoT was carved out as a standalone business from Vodafone Group in April 2024, backed by a ten-year Microsoft partnership built around “hyperscaling” the unit, and Skipper describes the resulting focus as near-absolute.
“We don’t try and get called into other areas… if you want all the other application things like mobile private networks and edge compute, you can buy those from other parts of Vodafone.”
Their current scale – 240 million-plus devices, 760-plus partner networks, 180-plus countries – gives the company considerable heft. The business model is a standard platform flywheel: narrower scope drives lower cost-to-serve, lower cost-to-serve drives margin, margin funds reinvestment in the one thing being sold, and the promise gets more reliable as the base gets bigger.
Note that Vodafone IoT leads the market by connection count in the West, not the global market. China Mobile holds roughly 44% of global cellular IoT connections against Vodafone’s 5%, mostly domestic volume rather than managed-service breadth. But within the markets Vodafone IoT actually competes in, the flywheel argument holds.
Skipper is proud of what Vodafone IoT has accomplished in the past decade, and justifiably so. However, there is some doubt about whether his insights into their strategy will hold good for other players in the industry, or whether it only works because Vodafone already had the scale to make it work before adopting it as a stated strategy.
Industry analysis of the 2026 IoT connectivity market offers some insights. Eseye’s CTO Ian Marsden has described most mobile operators as needing to “pick a direction”: either divest the IoT business to protect margins elsewhere – explicitly giving Vodafone’s IoT spin-out as the example – or partner with specialist connectivity providers because legacy platforms carry a cost-to-serve that “destroys margins when applied to low-revenue IoT devices.”
Meanwhile, other large players are moving in the opposite direction. AT&T has built its more recent IoT push around edge compute and AI network intelligence, launching an end-to-end solution stack with Cisco and. Cisco’s own 2026 rankings of IoT connectivity platforms explicitly frame “platform-led connectivity” – moving up the value chain into orchestration, security depth, and AI – as the thing that will differentiate providers “in a market where basic connectivity” is commoditizing.
That same ranking places Vodafone IoT as a leader specifically because of its “expanded single-pane-of-glass capabilities” and “accelerated AI integration” – meaning even Vodafone’s own external evaluators don’t describe it as staying purely narrow. It is investing in more capability – but it’s capability that manages complexity on behalf of the customer rather than capability customers interact with directly.
Reinvention
Skipper describes the SME segment as a long tail – maybe 20% of Vodafone IoT’s business, but hugely active now.
“They’re finding they can punch above their weight by having IoT. They can do things that they could have never done before without it.”
His example is that of a company running five cameras off IoT for a decade. That’s not a market where you’re going to make much money on device sales or even connections, and pricing on data is liable to be prohibitive to many businesses if the cameras run constantly. Instead, you have to return to a value proposition – being there as a support even if you’re never needed, and making sure things just work. That value has a price which is separate from the capacity or connectivity.
SME growth in IoT may well reflect a market that’s reaching greater maturity, as capabilities that were once difficult to use or unclear in their application become more accessible. A McKinsey survey found 51% of companies willing to invest in technologies like IoT specifically to gain competitive edge, and a 2026 European study found SMEs who are combining AI with IoT and data analytics saw growth potential rise by a further 21% over AI adoption alone.
The key thing is that a small company with no IT department can now run infrastructure that used to require enterprise-scale integration expertise. That said, some expertise may help – other research puts IoT project failure rates in the 50-60% range even where budget isn’t the constraint, which suggests the “punch above their weight” outcome depends on a level of hand-holding and clarity about what to do with the IoT devices.
Disruption may be coming to IoT, for all that it’s a mature market. As with many things these days, it seems like it’s AI-powered, but probably not in the way you’d assume.
His starting point is an economic rule of thumb used by the IoT industry: a hundred-dollar asset might historically justify a ten-dollar tracker. However, could you get a one-dollar tracker for a ten-dollar device? Or what about for a two-dollar piece of equipment?
There’s a new solution in town that might breaks that constraint, and it’s the use of AI to make cameras more intelligent as a sensor.
“With vision of course, you can look at so many different aspects of the same thing that you only need one sensor.” A camera and a good model to interpret the visual feed can substitute for the whole category of purpose-built trackers a given use case would otherwise need.
Interestingly, Skipper also suggests that there’s a whole new category of solution that intelligent machine vision might open up:
“IoT is very good when things are ordered. It’s not so good when things are disordered… the lovely thing about vision is that you can recognise and discriminate things on the fly, so you can turn disorder into order.”
He calls this “chaotic IoT” and gives two concrete examples – recycling streams, where “you never know what’s coming down the conveyor belt,” and airline galley trolleys, where the contents vary every flight – as categories that were effectively unsensorable before vision-based AI, because no fixed sensor configuration could anticipate what it would need to detect. Skipper suggests this may spark the next demand cycle for the whole IoT industry.
“We’ve seen sort of the first hump with 2G, the second hump was really the move to 4G… and then we see the hump which is going to be the big one, which is the movement towards AI and AIoT.”
There does appear to be substance to this. The AI camera market is forecast to grow from roughly $21.7 billion in 2025 to nearly $71 billion by 2030 – a 27% compound annual growth rate. Meanwhile edge AI, the processing layer that lets a camera make sense of what it sees without sending everything back to the cloud, is projected to nearly quadruple by 2033 (editor’s note – It’s already a contentious area, but that’s potentially a reason for telco players to get involved in this market).
Machine vision more broadly is a smaller but steadily growing market, expanding roughly 11% a year through 2030, driven by manufacturers building AI-powered edge-vision products for exactly the predictive-maintenance and quality-inspection use cases Skipper mentioned.
The Road More Travelled
What can we learn from the IoT evolutionary journey Skipper has been on in the past decade? There are some clear parallels to more greatly-hyped technologies today, which may give us some insights into their next few years.
That journey is as follows:
Step one: The hype outruns real deployment, and the “test the technology” phase is unglamorous and long. Skipper described IoT’s early phase as conversations with customers about whether it was even worth doing, followed by a painfully long industrialisation slog compared to the more exuberant forecasts.
Gartner’s 2026 data has agentic AI at the Peak of Inflated Expectations, with only 17% of organisations having deployed agents despite 60%+ planning to within two years, and separate reporting puts pilot-to-production failure rates as high as 89%, with 40% of agentic projects expected to be cancelled by 2027 on governance and cost-overrun grounds. That’s IoT’s “is this the right thing to do” phase, almost exactly.
Step two: Standards and a lowering cost curve, not technical capability, unlock scale. Skipper’s clearest example is that vision-as-sensor changes IoT’s economics because it removes the need for a dedicated tracker on every asset. Satellite is going through a similar unlock right now: 3GPP’s increasing TN-NTN integration means ordinary chipsets can talk to satellites without a proprietary modem. It commoditises the hardware that can what makes mass adoption possible.
AI hasn’t had its equivalent unlock yet. Inference costs and reliability (hallucination rates, governance overhead) are still the blocker cited across the Gartner data, not a lack of raw model capability. If the IoT pattern holds, AI’s next bump in adoption will only happen once somebody is able to correct that reliability piece.
Step three: Consolidation around scale and specialisation. IoT took roughly a decade to reach the point where operators had to “pick a lane” as per Eseye’s report. Satellite is consolidating rapidly now, as different kinds of economics come to the fore in a market that’s simultaneously decades old and in some ways brand new. Meanwhile AI’s consolidation is happening at the infrastructure layer where a handful of hyperscalers own the compute and frontier models. IN contrast the telecom application layer is still undisciplined, with operators variously and simultaneously chasing AI as a network solution, a customer solution, a sellable product and a platform play. This may be some players replicating the “reach up too high into areas where you lack knowledge” mistake Skipper described as IoT’s classic error.
Step four: Governance and standards lag deployment ambition, and vendors fill the gap with proprietary frameworks. Satellite has a version of this; as recently reported, the GSMA’s work on international NTN regulation highlights that “several nations [have] yet to formalize NTN frameworks,” and cross-border spectrum-sharing disputes are cited as a current business risk.
An industry or vendor-published “framework” is a signal that the governance or policy infrastructure a technology needs doesn’t exist yet rather than evidence of a gap being closed. However, a focus on governance is a good hint at a more mature market, or at least one that’s reached a certain size worth worrying about.
Eventually, the pitch stops being about the technology itself, which simply becomes an enabler. This is already visible in satellite; “no app, no setup, it just works” is exactly how direct-to-cell is being marketed to consumers. AI has a long way to go. “AI-powered” is a selling point right now precisely because AI hasn’t yet become simply understood as a default part of the infrastructure.
There are some clear differences in the nature of these technological beasts, but Skipper’s journey gives us some useful markers as well as a final destination where the business is all about the customer, no matter how humble.
