Private 5G has long carried a reputation for being heavy, complex, and slow to deploy. But it can be a lightweight and scalable foundation for AI in mission-critical industrial operations, as well – provided it is designed properly. As enterprises race ahead with AI investments, the company says the missing ingredient is often the connectivity layer itself – and that private 5G, done right, is ready to fill the gap.
In sum – what to know:
Heavy investment – industrial firms are investing in AI, robotics, and vision systems – but many discover too late that legacy connectivity can’t deliver the control or determinism these applications demand.
Cloud orchestration – Celona’s AerFlex architecture aims to strip private 5G of its complexity, enabling multi-site deployments with on-premise access points with local breakout and cloud-based control-plane functions.
New partnerships – new partnerships around AI orchestration, modular edge compute, and managed services mean Celona is positioning private 5G as an integrated infrastructure layer for physical AI.
The predominant picture of private 5G, perhaps, is one of complex technical deployments at sprawling industrial venues – with heavy gear, deep integration, and long ramp-up cycles. But US-based specialist Celona thinks private 5G can be perfectly lightweight, flexible, and accessible, given the right architecture in the right environment – and become the new connectivity substrate for AI, edge computing, and mission-critical operations. But too often, the message has not got through, the company says.
It has plenty of examples where clients have only been in touch after investing in AI infrastructure – in compute systems, camera installations, and IoT sensors. “Customers assume the right connectivity is already there… And then when they actually start to deploy [AI applications], they can’t take advantage of them… The only way to upload data has been to go back to the office where there is connectivity. Where they needed it, they did not have it,” says Puneet Shetty, vice president of product at Celona, speaking during a session at Industrial Wireless Forum
On the phone, a week later, he goes further. “A lot of the investments are happening in AI but one of the things that gets forgotten is the connectivity layer.” Many industrial players find that “their Wi-Fi is built for their IT functions, rather than OT functions.” For AI, robotics, IoT or video analytics, that kind of legacy connectivity “does not support” the flexibility, reliability, or determinism those applications need, he says. In Shetty’s view – and increasingly in the view of the wider industry – private 5G is emerging as the optimal connectivity layer for “physical AI”.
But there is a perceived trade-off, he suggests, between high performance and easy usability.
Missing layer
Wi-Fi is easy and familiar, the story goes, but lacks the ultra reliability, deterministic latency, high security, and seamless mobility that mission-critical operations demand. Public cellular networks, on the other hand, offer scale, but no control, and often no guarantees on latency or performance. As companies embed AI-geared automation and robotics, the limitations of legacy networks are clear. As above, many organisations are rushing to buy compute and camera systems only to discover their old networks can’t handle the load, or deliver the control.
Private 5G is the answer, says Shetty. Where it has gained traction, its adoption has not been driven by flashy generative or agentic AI use cases, but by more prosaic – yet transformative – “physical AI” deployments: camera vision systems, coordinated robotics, connected-worker tools. For many industrial enterprises, just the move from paper forms to digital workflows over a robust wireless network already represents a meaningful transformation. But early private 5G solutions have been cumbersome to deploy and manage, and enterprises have struggled.
Early generations of private networks were complex and costly to deploy – typically requiring on-site servers, dedicated controllers, and significant technical effort. Shetty says: “Three years ago, we were still in POC hell – where only a few trials expanded into commercial deployments, and the discussion was mostly a technical one. Whereas projects are now very focused on specific paint points, and, mostly, land on a couple of options: Nokia, Ericsson, and us.”
Celona’s anticipated rollout in 2025 fell some way short of its expectations, he acknowledges. “We anticipated a certain level of growth, and accelerated our go-to-market. But there has been uncertainty in the market, especially in the industrial space, and so our rollout plans have not necessarily panned out. Where we expected customers to scale to 20, 30, 50 sites during the year, more commonly, they only deployed at 20 to 25 percent of their targets. A lot of it has moved to next year.”
Market realities
All through, Celona’s approach has never changed, he says – to make private 5G as simple as possible for enterprises, without losing any of the technology’s unique capabilities. In August, the company launched AerFlex, a cloud-controlled single-unit private 5G solution, which splits the core network software between the access point (AP) hardware and the cloud, and eliminates the need for on-site servers or complex integrations. The idea is to unlock easy multi-site scalability in the private 5G market.
Enterprises need to only deploy access points, like they do with Wi-Fi; orchestration, control and management live in the cloud. Rollouts that once took weeks – involving racks, cabling, controllers – can be completed in hours, even across multiple sites. The burden on IT/OT teams is reduced, lowering the barrier to expansion. Thanks to Celona’s unified software stack and intelligent splitting of functions (radio, core, orchestration) between APs and cloud, the solution supports high-end physical AI for Industry 4.0 without compromising data compliance or performance SLAs.
Shetty says: “Ours is already an enterprise-friendly solution, versus some of the bigger vendors, and we own the whole stack, versus some of the Frankenstein solutions from some of smaller companies. So others are offering the radio or core, or the orchestration; but that is not how enterprises consume connectivity. For Wi-Fi, they are a Cisco or a Juniper shop. They will consume private 5G in the same way. We are bringing the whole puzzle. And Aerflex goes on top of our existing portfolio – so enterprises can deploy private 5G with just a radio.”
He adds: “The idea of moving components to the cloud is not new. What is unique, again, is that we own the stack, and can split it as we want. Some have moved the entire core to the cloud, which is a broken concept because industrial enterprises will not accept that the data plane is in the cloud; it needs to run on-premise. But we can choose what goes where – so the data plane is instantiated on an access point, which allows us to do local traffic breakout on certain sites, while the data plane can reside on an edge appliance at bigger campuses.”
He adds: “That flexibility is unique to Celona versus a lot of these other vendors.”
Integrated design
Celona isn’t doing it alone. In 2025, the firm has forged various strategic partnerships that signal a shift – where private 5G is not just about radio equipment and orchestration software, but a part of an integrated infrastructure proposition to drive AI automation and operations at the edge. It has just announced an integration with Sutherland, combining Sutherland’s AI orchestration and management stack with Celona’s private 5G system. The result is a “self-managing, intent-based network”, it says, to reduce human intervention while boosting uptime and resilience.
Celona also has a new collaboration with Armada to deliver “digital oases” – modular edge-compute sites integrating private 5G, satellite backhaul, and pre-configured AI models. The aim is to deploy autonomous AI infrastructure “anywhere in the world” in as little as 48 hours. The solution is geared for harsh or remote environments, out of the way of easily cabled cloud access. Celona has a new channel deal with Hughes Network Systems, as well, to offer a ‘turnkey’ managed private 5G service; they are targeting enterprises without in-house cellular or network expertise.
The offering spans site design, deployment, 24/7 monitoring, and support – switching private 5G from a high-maintenance DIY project into a managed service. Again, these moves reflect the broader shift – that private 5G is no longer a stand-alone connectivity play, but a core layer in the advancing enterprise stack. The strategy reflects larger structural shifts in the private wireless market. Market watchers are bullish on enterprise demand for superior connectivity – especially where AI and IoT are geared for real-time automation projects.
But industrial enterprises are cautious and pragmatic, by their nature – even if geo-politics and macro-economics are removed from the equation. Shetty says: “They are very pragmatic, of course, and probably a little suspicious of what AI will do for them. But it is a mixed bag; some manufacturing customers and many industrial automation vendors have gone all-in on AI.” He cites Celona’s work with US process manufacturing giant Cargill, a business with around 160,00 employees and $165 billion in turnover, and a presence in 70 countries.
“You ask Cargill, and it is not waiting for an ROI; it is not waiting for a killer application to appear. Its view is that industrial enterprises should understand that, at some point, whether in 12 months or three years, they will need private 5G. It is evaluating Wi-Fi versus 5G, but it knows it will have AGVs and AMRs in a few years, so it is preparing for that – and upgrading to private 5G. But really, you have a mix of customers, between those who want to see the ROI on their applications, and those who know everything is going to change, and they need to be ready.”
Long tail
Either way, the vendor market is required to simplify private 5G deployments – both for early adopters, and the long tail, including small and medium-sized firms (SMEs). AP-only cloud-controlled architectures, of the type Celona is popularising with its AerFlex system, are a step in that direction; so too is its work with partners to enable managed services and build integrated edge systems. And its holistic take, especially for a start-up, is rather unique – core, radio, and orchestration, like only Nokia and Ericsson, of the big Western players, properly do.
As such, Celona is doing as much as anyone to shift the perception of private 5G – from optional infrastructure to something more: a fundamental enabler of modern enterprise digital transformation. For companies investing in AI-driven automation and robotics across distributed edge setups, such an integrated and managed service offer, rather than a patched-up cut-and-shut, appears like a good thing. Its vision resonates especially well in industrial contexts, says Shetty – in factories, ports, warehouses, mines, energy infrastructure.
He sums up: “As folks look at agentic AI now, where an associate on the floor talks to an agent about a sensor reading, then the conversation and application suite on private 5G is expanding. Those applications will have to be, especially in an industrial environment, enabled by a very strong and secure connectivity layer. And private 5G will be the answer to that. Generative and agentic AI are not big drivers for private 5G, yet, but private 5G will be the foundation for these things, as enterprise stack use cases on top.”
