Private 5G and generative AI are redefining industrial connectivity and intelligence – but do they truly depend on one another? This discussion explores how these enabling technologies overlap, complement, and diverge in the enterprise stack – as manufacturers and operators push toward a more autonomous, data-driven future.
Note this article comprises the introduction and opening section to a new editorial report from RCR Wireless about private 5G and generative AI. It takes a write-up of a webinar session on the same subject as its base (and also copies its header), and colours it with various other related interviews and articles as a start-point for a longer report discussion. A more recent discussion with John Deere might have also made the cut, had it been conducted sooner. The full full report can be found here, and will be serialised online over the coming weeks.
In sum – what to know:
Parallel enablers – private 5G and generative AI are both major drivers in Industry 4.0, enabling automation, intelligence, and flexibility, but they do not directly depend on one another.
Building realism – private 5G adoption is accelerating, with enterprises to spend $10bn by 2030; AI investment, and especially gen AI, remains a small but fast-growing part of industrial strategies.
Some convergence – BMW, Volkswagen, Airbus, and Newmont are advancing AI and 5G deployments, though mostly for efficiency and safety, rather than generative AI experimentation.
Introduction
There is an argument, perhaps, that generative AI and private 5G are the two biggest hype stories in industrial tech right now – going up and down either side of the curve respectively. Maybe that is a telco tale; but private 5G has built a genuine head of steam in Industry 4.0 quarters since the start of the year, to the point large-sized commercial deployments are announced every day, and generative AI is a total phenomenon, of course, going way beyond factory lines, into homes, offices, and the corridors of global power. But the high point for AI, like with 5G, is inside enterprises – where it is required to be customized and dynamic, and where it promises to change the economy.
All of which is why RCR Wireless sought to combine these topics in this report, and in an attendant webinar last month – to consider how they mesh, and even if they mesh, in service of Industry 4.0. So is there a connection between them – really? The panel scratched its chin, and the audience wondered out loud – and the response was kind of overwhelming: no; but also yes, everyone said. The point is that 5G and AI – whether private and generative, or not – are what engineers and marketers like to call ‘enabling technologies’. One does not beget the other, very clearly; you don’t need 5G for any kind of AI, unless perhaps you are out in the sticks, and the sums work.
Just to jump around: there was a similar discussion during a private-5G panel at FutureNet World in London a few months back (in May). Raise your hand if you think private 5G needs AI. It was a question for the audience, and the London panel never quite got past it. Not because only three hands (in a graveyard shift) went up, but because the question is two-sided, at least – and mobile operators BT, KPN, and Telenor argued the toss. Does it need AI? Of course it doesn’t need AI – if you just want to automate cabled equipment, and AI is not more than pattern-matching ML to drive telegraphed operational efficiencies. But private 5G is also being sold in decent numbers for such ends.
Today, private 5G is in a “good spot”, said Tommy Björkberg, vice president of network and cloud at KPN, in London. “There’s more to be done, but there are really interesting use cases that deliver value for enterprises.” And clearly, lots of those industrial use cases are interesting and creative because they use AI in one form or another. So need might be too-strong a word, but 5G and AI get along pretty well. Because, as above, they are horizontal ‘enabling’ technologies, even when they are privatised and customised for vertical industries; one goes above the other in the classic tech stack – and their parallel narratives converge at points, as use cases require.
Because AI works on 5G, too, of course – maybe less well than on fibre, but then 5G brings mobility, which brings flexibility, which enables certain applications; but better than on Wi-Fi, probably, except that Wi-Fi is well-understood, and getting better. In some cases, where extensive cabling or coverage is required, 5G is just cheaper, too. So, there is plenty of crossover; it’s just not a direct one-plus-one relationship, necessarily. But maybe it is, as well – as we will discover.
Part 1 | Industrial momentum and market reality
The hype around private 5G and AI is being tempered by the realities of enterprise demand. Industrial leaders like Michelin, Axa, and Thames Freeport are discovering that connectivity is the foundation for intelligent automation – and that value begins with practical deployment, not futuristic promises.
In the RCR webinar session, James Moar, principal analyst at Kaleido Intelligence, quotes a company forecast that over $10 billion will be spent on private networks by 2030. Which sounds like a lot, and probably is – more than three times what is being spent on them today. Indeed, projections for the private 5G market all spiral upwards, no matter which analyst firm you ask. Momentum has “extended” through the first half of 2025, says Dell’Oro Group; revenue growth (for private 4G/5G RAN sales) will be about 20 percent up on 2024 (on 40 percent growth on 2023) by the end of the year. It describes the market as a “still-untapped, high-growth opportunity”; like a bulging niche discipline.
Private 4G/5G contributes a “mid-single-digit share” of total global RAN sales in 2025, reckons Dell’Oro Group – on a course between a “low single-digit share in 2022” and a higher single-digit share by the end of the decade. Sales of public 4G/5G RAN systems are “flattish”, meanwhile – and will presumably remain so, through stop-start 5G SA macro upgrades, until operators embrace the 6G era. More than just a “good spot”, private networks represent a “bright spot” in the whole telecoms market, says Dell’Oro Group. This is because enterprises are suddenly thinking about the criticality of their networks – with a view to some kind of corporate AI-sprung promised land.
Which is the message from insurance firm Axa, tyre manufacturer Michelin, and fashion house Prada, sharing a stage at a Verizon Business partner event in Paris in October. Critical connectivity is, suddenly, a key boardroom concern in global enterprises, they say. “The discussion has completely changed,” says Prada. “Three years ago, the network was only talked about in the boardroom when it wasn’t working.” Same, says Michelin, adding: “As a data-driven company, the transfer of data has become key.” Same, says Axa: “The network is the very [foundation] of our services.” Daniel Lawson, senior vice president for global solutions at Verizon Business, cuts to the chase.
“All these AI data centers [will] be very expensive paper weights if they don’t connect to anything,” he says. None of which is about private 5G, specifically; these companies are taking network services of other kinds (mostly SD-WAN for fibre, maybe some IoT) from Verizon Business. But their comments capture the discussion, very well, and it is boiled down to its essence in a subsequent panel in Paris with Thames Freeport – which is deploying six private 5G networks at four logistics hubs (an old port, a new port, a big warehouse, and a manufacturing plant) in a new designated UK ‘economic zone’ along a stretch of industrial land on the north bank of the river in London.
Asked why private 5G was picked as the first part of the Thames Freeport puzzle, Tom White, director of innovation for the project, responds: “We are looking to bring high-tech modern industries to provide high-productivity jobs. And we stripped that back to the things we could invest in that those industries needed… to move materials around more cheaply, to have more resilient systems, to deploy more AI into manufacturing.” Private 5G provides the “foundational connectivity” to build such applications, he explains. What is not covered in Paris – and has not been covered anywhere (and won’t be in these pages, yet) – is its plan to host new AI data centres at the site, as well.
The point is that this 5G/AI linkage is close, and getting closer. But AI is a broad church, too. The global industrial AI market was worth $43.6 billion in 2024, according to IoT Analytics, with compound growth (CAGR) pegged at 23 percent per annum through 2030 – when it is expected to be worth $153.9 billion. The growth is super-charged by the buzz about generative AI. But two things: spending on AI represents just 0.1 percent of corporate industrial revenue, and generative AI, so hyped, represents less than five percent of AI projects in the industrial market – and these are mostly IT-grade pursuits, in operations and service support (documentation querying and troubleshooting).
“Manufacturing rollouts have been driven by industrial software vendors in copilots,” writes IoT Analytics. But generative AI is also being deployed in code generation for OT and embedded assets, and will feature more heavily in R&D (product discovery), design (generative design), engineering (gathering requirements), and field service (guided maintenance), says IoT Analytics – to the point it comprises a quarter of industrial AI projects by 2030. When IoT Analytics polled execs at big manufacturers in 2021, AI was hardly on the radar, rarely appearing in more than “ad-hoc exploratory projects”; most firms now have a CEO-driven AI strategy (as per the Prada comments), it says.
But a poll by Kaleido Intelligence finds the opposite: that most enterprises are not deploying 5G because they want AI, let alone generative AI. They want “basic benefits” from private 5G, says Moar on the RCR panel. “They want security, reliability, and privacy; the most exotic thing is guaranteed quality-of-service. Which is a big feature for network slicing. But that is about as techy as you get. So to drive this market, [there should be a] focus on basic things, and to build from there.” Which is a familiar lesson, of course – that enterprises don’t buy tech, only solutions. When the rubber hits the road, anyone selling an industrial version of generative AI will face the same reality-check.
A panel of reflective vendors will conclude, in a couple of years, that the initial problem with AI was that it was sold as a rebel tech without a cause. But there are differences in these samples: Kaleido Intelligence surveyed firms of “all sizes, down to small enterprises”; IoT Analytics polled the big guns in global manufacturing. “You’ll get something very different if you ask an international automaker,” says Moar. (Discussion about how AI will cleave open a digital divide in Industry 4.0 are for another day.) IoT Analytics sees a major shift, where boardroom AI strategies are “vision-driven, supported by governance frameworks, performance targets, integration with broader objectives”.
Indeed, just look at the latest updates from Volkswagen and BMW: the former has said it will invest €1 billion ($1.17bn) in AI-related industrial tech by 2030 to boost vehicle development, industrial applications, and IT infrastructure; the latter has just opened a new “fully-connected” AI car factory in Debrecen, Hungary, billed as the “most innovative” site in its global factory network (of 30 production facilities). Volkswagen wants “no process without AI”, it has said. BMW is squarely focused on AI in its production lines and logistics setups. Neither of them mention private 5G in their announcements, and neither have ever said very much about it, anyway; but both are using it.
But then, they don’t say a whole lot about production-ready generative AI, either – reflecting either its rather niche application as a support function or the unresolved risk and trust issues that go with it (per the IoT Analytics stats). BMW, say, talks-up AI as a solution for a “wide range of quality checks… on the production line”, but never mentions generative AI. Its AI “centrepiece”, deployed at its smartest plants, is an in-house cloud platform called Artificial Intelligence Quality Next (AIQX), developed with acoustic sensors, microphones, and cameras to run automated quality checks on conveyor lines – capturing data about each vehicle’s position, componentry, and finishes.
The data is analysed for anomalies on a trained model in a cloud-back-end, and returned as alarms for staff about missing or misassembled parts, and other assembly-line defects. But again, it’s not a generative support model – in the first instance anyway. Take also the cases of Airbus and Newmont Corporation, both pin-ups for private 5G in their respective industries. Airbus has just deployed private 5G at plants in Hamburg and Toulouse, taking its footprint to five, with plans for new networks in the UK, Spain, and the US. Newmont, the only gold producer in the S&P 500 Index, has 5G at its Cadia mine in New South Wales, and plans to deploy at 14 mines on four continents.
Both are using Ericsson gear. Their use-case rosters are virtually the same, and make no mention of generative AI: variously, digital twins for design simulation, augmented/virtual reality (AR/VR) for worker support, automated vehicles and robots (AGVs and AMRs) for automated production, plus enhanced IoT cases like asset tracking and predictive maintenance. Airbus says: “Private 5G forms the backbone of [our] strategic transformation projects, enabling high-value industrial use cases such as IoT integration, intelligent management of critical equipment, real-time quality control, and collaborative robotics.” How much is AI an end-goal and a motivation?
It’s a silly question, perhaps, given the comments about “non-techy” drivers; but it is a direct one, asked of Chris Twaddle, director of networks in Newmont. Does the trend for generative AI inform Newmont’s 5G strategy? He responds: “Our main motivation is to keep people safe and improve efficiency. While we use AI across the business in various ways, it hasn’t been a key driver of our 5G rollout. But the concept of site-wide high-speed wireless is really attractive to teams exploring how to increase AI in our operations. If you think about computer vision, 5G gives the ability to quickly and efficiently deploy or move a camera to where computer vision is needed to add value.”
Which is maybe the best quote in this whole discussion. But another thing quickly, for context: 5G is getting better, especially in private networks. As it stands, lots of private 5G is actually private LTE (4G) – in usage, in practice. However, the old 80/20 rule will be reversed by 2030, says Kaleido Intelligence, when 5G SA infrastructure will be more widely deployed in industrial enterprises. It will be more advanced than in public networks, too; private 5G systems already feature slicing and RedCap, notes Moar. “They were initially built for mission-critical connectivity, but we’re now seeing use cases move beyond that, and other things get attached – so usage is kind of snowballing.”
He adds: “Once the initial ROI is proved, it rapidly becomes something that everyone starts to use – for everything.” And so, suddenly, AI abounds in better networks, and generative AI affords a new way for non-specialist 5G operators to understand and manage both the networks themselves, and all of the production-line AI pyrotechnics that they connect. Right?
To be continued.
The full editorial report, featuring this article, can be found here, or by clicking on the image below.
