In short: The "Wi-Fi or cellular?" argument on the factory floor is usually framed as a religious war. It isn't. There is a set of concrete questions you can walk through — about motion, density, determinism, handover, spectrum, and operating model — that will tell you which technology fits your plant. This guide lays them out in the order a plant manager would actually ask them.
Key Takeaways
- The deciding factor is almost never bandwidth — it's whether anything on your floor moves, whether anything on your floor has a real-time control loop, and whether you can tolerate a handful of seconds of uncertainty per device per day
- Wi-Fi 6 is the right answer more often than the 5G industry admits — if your automation is mostly stationary and your devices are tolerant, you can run a lot of Industry 4.0 on a well-designed wireless LAN
- Private 5G wins on three specific workloads — moving robots, safety-critical control, and multi-service consolidation — and loses to Wi-Fi 6 on cost, familiarity, and installation time everywhere else
In a nutshell
The Argument That Isn't Really An Argument
Walk into any plant and ask how the shop floor will be connected in five years and you will get one of two confident answers. One camp will tell you that Wi-Fi 6 and 6E have closed the gap on cellular, that your IT team already knows how to run it, and that private 5G is a telco sales pitch looking for a problem. The other camp will tell you that Wi-Fi is fundamentally unsuitable for industrial automation, that cellular is the only deterministic wireless technology, and that anyone still putting access points on the ceiling of a factory in 2026 is setting themselves up to rip it out again in 2028.
They are both partly right. They are arguing about different plants.
The useful question is not "which technology is better." It's "what is on my floor, what is going to be on my floor, and what do those devices actually need from the network?" That decomposes into a small number of questions that a plant engineer can answer without a telecoms degree.
Question 1: Does Anything Move?
This is the single biggest discriminator and it gets asked last more often than it should.
Wi-Fi was designed for roaming laptops, which is not the same as designing for moving robots. When an AMR or an AGV crosses from the coverage cell of one access point to the coverage cell of the next, the handover has to happen during movement — and Wi-Fi handover is a client-side decision, made by the device firmware, without the help of the network. In practice, that translates into pauses of anywhere between 200 milliseconds and a couple of seconds while the client re-authenticates and re-associates. For a human holding a scanner gun, that pause is invisible. For a mobile robot executing a path-planning command, it is a stop.
Cellular networks were designed around mobile phones, and the network is actively in charge of handover. Modern 5G hand-offs complete in tens of milliseconds, predictably, whether the device is a phone, an AGV, or a forklift. If anything meaningful on your floor moves — logistics robots, autonomous forklifts, mobile cobots, rolling inspection stations, even hand-pushed trolleys with heavy tablets — the calculus leans cellular almost regardless of everything else.
If nothing meaningful moves, this question disappears and Wi-Fi stays in play.
Question 2: How Tight Is Your Control Loop?
The second question is about determinism. Most devices on a factory floor are actually best-effort consumers of the network: a vibration sensor sampling at 100 Hz and uploading a 30-second window every minute does not care if a packet is delayed by 50 milliseconds. An MES terminal displaying work orders doesn't care. A pick-to-light system doesn't care. A camera feeding a batch-mode AI inspector running at the end of shift doesn't care.
Some devices do care. A servo drive on a printed-circuit-board line running a synchronised multi-axis move cares about jitter at the sub-millisecond level — but it will be on a wired fieldbus anyway, because no wireless technology comfortably does that. A cobot coordinating with a human operator in a shared workspace cares at the 5–10 millisecond level. A tool exchange on a mobile machining cell cares at the tens-of-milliseconds level. An AR maintenance headset cares at the 20 millisecond level, but mostly because humans feel lag viscerally. An AI inspection camera feeding a live classifier that must reject a part before it leaves the station cares at the hundreds-of-milliseconds level.
Wi-Fi 6 can serve many of these workloads well. Where it struggles is in guaranteeing that it will serve them consistently as the floor gets busier, because Wi-Fi is contention-based: when the medium is crowded, access is fought for. Private 5G is scheduled: the base station tells each device when to transmit. The practical effect is that 5G's worst case stays close to its average case, which is exactly what a safety or control engineer wants.
If you have any real-time control loops on the floor, or if you expect to add any, the answer is cellular. If you have none, Wi-Fi 6 is a legitimate option.
Question 3: How Much RF Noise Does Your Process Make?
Factories are unfriendly environments for unlicensed spectrum. Variable-frequency drives, welding, induction heating, and large motors all dump broadband noise into the 2.4 and 5 GHz bands. Metal surfaces create multipath reflections that are hard to predict. Forklift bodies move the interference pattern around during a shift. A Wi-Fi survey taken on a Monday morning when the line is cold is not the same as the RF environment during a Wednesday afternoon production peak.
Wi-Fi 6E opens up 6 GHz, which is cleaner today but will not stay clean as adoption grows. Private 5G operates in licensed or shared-licensed spectrum — in the UK, the Ofcom Shared Access Licence framework makes frequencies in the 1800 MHz, 2300 MHz, 3.8–4.2 GHz, and 26 GHz bands available to enterprises at low cost, with the legal protection that nobody else is allowed to transmit on them inside your polygon. That protection is the thing.
If your process is electromagnetically hostile, licensed cellular is dramatically more robust. Wi-Fi can be engineered to survive it, but every change to the floor is an opportunity for the survey to become obsolete.
Question 4: How Many Devices Are You Running?
Wi-Fi 6 handles device density well for a wireless LAN — a well-designed access point can host a few hundred clients in practice. Private 5G handles it several orders of magnitude better, up to the million-devices-per-square-kilometre figure that the standard was designed around.
For most plants this question doesn't matter. If you have thirty-odd AMRs, a hundred handhelds, and two hundred fixed sensors, either technology is fine. But if you're building a sensor-dense plant around bearing monitors, temperature probes, flow meters, power meters, and environmental sensors at every workstation, the numbers climb quickly into thousands. At that point cellular is materially cheaper per device and simpler to design.
Question 5: Who Is Going to Run It?
This is the question that hands Wi-Fi its strongest win.
Your IT team has been running Wi-Fi since the early 2000s. They know the vendors, the tooling, the monitoring, the troubleshooting steps. Adding Wi-Fi 6 to the shop floor is incremental to an existing skill set. Private 5G is a genuinely new operating model — the technology looks more like a small telco than an enterprise network — and the staffing implications are real.
This is where a managed private 5G network changes the arithmetic. If somebody else is running the core, the radios, the spectrum licence, the monitoring, and the patching, the operating-model objection disappears and you get to evaluate 5G purely on the technical questions above. If you plan to run it yourself, budget for an uplift in your networks team and be honest with yourself about whether you can hire for it.
A Shortcut You Can Use Today
If you want a single-page rule of thumb, this is the one we give plant managers on a first call.
Choose Wi-Fi 6 / 6E if: your automation is mostly stationary, your control loops live on wired fieldbuses, your device counts are modest, your RF environment is tolerable, and your IT team already runs a credible enterprise wireless LAN.
Choose Private 5G (usually managed) if: any meaningful part of your automation moves, you expect multiple real-time workloads to share one network, your process generates enough RF noise to make Wi-Fi surveys feel like guesswork, you want one wireless layer for OT and IT traffic with guaranteed separation, or your plant is large enough that access-point density starts to dominate the cabling budget.
In practice the real answer at most large plants turns out to be both — Wi-Fi 6 for office traffic, handhelds, and non-critical sensors; private 5G for the robots, the cameras doing live inspection, and the safety-critical workloads. The two technologies don't compete as much as the trade press suggests. They do different jobs on the same floor.
UK Context
The UK's Made Smarter programme continues to fund digital adoption in manufacturing, with several regional hubs running direct comparisons between Wi-Fi 6 and private cellular on live lines. Ford's Halewood facility has publicly discussed its connected-factory direction as part of its EV transition. Nissan Sunderland's EV39 expansion has involved significant shop-floor wireless rework. Airbus Broughton has been an early testbed for industrial cellular in aerospace. Each of these programmes has ended up with a mix — not a winner-takes-all outcome — which is the most reliable signal about where this is heading.
Plant managers trying to make the call today should not feel obliged to pick a religion. The factory floor of 2030 will have both technologies on it. The useful decision is which workload goes where — and that decision is engineering, not faith.
If you're designing the wireless layer for a new or retooled plant and want an honest second opinion on the mix, get in touch. Read more on our manufacturing sector page.