Plant managers keep asking us the same question: "We already have Wi-Fi everywhere — why do we need private 5G?" The short answer is: you might not. But if your automation layer has any device that requires deterministic latency, guaranteed handover continuity, or SLA-backed uptime, the answer changes fast.
This article is not an advertisement for private 5G. It's a decision framework. Both technologies are mature enough to work in industrial environments — the choice comes down to what your specific device mix actually needs from the radio layer.
The core difference: contention vs. guaranteed bit rate
Wi-Fi operates on contention-based medium access. Every device on an AP competes for airtime using CSMA/CA. Under light load, this works fine. Under dense device populations — 80 AGVs, 200 sensors, 40 operator tablets all sharing the same spectrum — contention increases, retransmissions climb, and effective latency becomes unpredictable. Wi-Fi 6E (802.11ax) improves this significantly with OFDMA and MU-MIMO, but the contention model is still there.
Private 5G NR uses scheduled access. The gNodeB allocates resource blocks (RBs) to devices in each 1 ms transmission time interval (TTI). Devices do not contend — they transmit when the scheduler tells them to. This is not a theoretical advantage. It means you can configure a guaranteed bit rate (GBR) for a device class, and the scheduler enforces it even when the cell is loaded. An AGV's control channel gets its 512 kbps at 10 ms latency whether or not 50 other devices are active on the same cell. The 5G NR scheduler does not negotiate — it enforces the GBR allocation as a hard reservation in the resource block grid.
Where Wi-Fi 6E is the right answer
Be direct: Wi-Fi 6E is the right choice for most general-purpose industrial connectivity. If your use case is operator tablets, barcode scanners, IP cameras, and non-critical sensor telemetry, Wi-Fi 6E delivers excellent throughput at much lower deployment and operational cost. The access point density you already have for Wi-Fi 5 likely needs only modest uplift for Wi-Fi 6E.
Wi-Fi 6E also handles high-throughput bursty workloads well — machine vision cameras pulling frames across the network, MES terminals loading large data sets. The 6 GHz band gives you clean spectrum away from legacy 2.4/5 GHz congestion. Target spectral efficiency of 9–11 bits/s/Hz under ideal conditions means a single 80 MHz channel can sustain well above 500 Mbps aggregate for a dense tablet population.
We're not saying Wi-Fi 6E fails in industrial environments — in most of them, it works well for the traffic it was designed for. The case against Wi-Fi for automation control is not that it fails constantly; it's that it fails unpredictably. A single AP firmware update in the wrong maintenance window, or an adjacent AP's channel reassignment during peak floor activity, can spike your floor-average latency from 8 ms to 40 ms. For general data traffic, that's invisible. For an AGV at 2 m/s, that 32 ms gap is 64 mm of uncontrolled motion on a device that was supposed to be stopped.
Where private 5G NR is the right answer
Three categories of devices should push you toward private cellular:
Safety-critical motion control. AGVs, AMRs, and robotic arms executing moves under network-dependent control loops need latency guarantees your Wi-Fi infrastructure cannot contractually provide. 3GPP defines 5QI 82 and 5QI 83 specifically for discrete automation — these QoS classes specify a packet delay budget of 10 ms and a packet error rate target of 10⁻⁴. No equivalent QoS framework exists in 802.11ax. You can prioritize traffic in Wi-Fi, but you cannot guarantee it.
Handover-sensitive mobile assets. A forklift or AMR moving across a large DC floor transitions between APs continuously. Wi-Fi roaming via 802.11r (fast BSS transition) works, but handover latency is typically 50–100 ms and depends heavily on AP vendor implementation and client driver quality. 5G NR handover is managed by the RAN scheduler and completes in under 20 ms under normal conditions — and for intra-gNodeB handovers, often in 10–15 ms. For devices with open control loops during motion, that difference is the margin between a clean transition and an emergency stop.
OT/IT network separation. Private cellular gives you spectrum-level isolation from your IT Wi-Fi infrastructure. A broadcast storm on your enterprise WLAN cannot reach your AGV control channel if they're running on separate spectrum. In regulated industries or OT networks with strict IEC 62443 segmentation requirements, this is a compliance argument, not just a performance argument. The separation is at the physical layer — not just a VLAN boundary that a misconfigured switch can bypass.
A typical hybrid deployment: what it looks like in practice
Consider a 250,000 sq ft automotive sub-assembly plant with 60 AGVs on a pallet transport loop, 400 operator tablets, and 180 fixed sensors reporting to a SCADA concentrator. The right architecture for this facility is almost certainly not "all private 5G" or "all Wi-Fi 6E."
Private 5G on CBRS covers the AGV travel lanes, the SCADA polling paths, and the inter-zone corridors where handover density is highest. Wi-Fi 6E covers the assembly benches, operator access points, and office-adjacent zones where tablet density is high and latency requirements are relaxed. The two networks coexist without spectrum overlap: CBRS at 3.5 GHz, Wi-Fi 6E at 5/6 GHz. Interference between them is negligible.
The friction point in hybrid deployments is not the RF — it's the management layer. The CBRS cells have their own NMS. The Wi-Fi has another. The AGVs traverse both coverage domains. Without a control plane that aggregates device-level SLA status across both networks, your network team is doing triage in two separate tools every time an AGV stops unexpectedly.
A decision framework for your device inventory
Before making a technology recommendation, run your device inventory through these questions:
- Latency requirement: Does any device require guaranteed sub-20 ms latency? If yes, private 5G is the only architecture that can contractually enforce it.
- Roaming density: Do any mobile assets traverse more than 3 AP boundaries in a single shift? Frequent roaming is a handover failure risk on Wi-Fi.
- Safety classification: Are any devices in scope for IEC 61508 SIL or ISO 13849 PLd/e? Safety classification typically requires deterministic communication paths with documented worst-case latency.
- OT segmentation: Is spectrum-level isolation from enterprise IT a compliance requirement? VLAN-based segmentation on Wi-Fi is logical, not physical.
- RF environment: Do you have plasma cutters, induction heaters, or brushed DC crane motors? These generate broadband interference that degrades 2.4/5 GHz Wi-Fi more severely than 3.5 GHz CBRS, and the SAS priority structure gives your CBRS deployment a level of interference accountability that unlicensed Wi-Fi bands cannot provide.
If none of those apply, extend your Wi-Fi 6E deployment. If two or more apply, scope a private 5G pilot for the specific device classes that need it — not for the entire campus.
What not to do
The most common mistake is deploying private 5G at full campus scale before validating the management layer. You can have excellent CBRS radios, a solid gNodeB, and a UPF in the right place — and still fail to deliver SLA if you have no mechanism to enforce QoS policy across device classes or detect when a cell drifts below your SINR threshold. The RAN hardware is not the whole stack. SLA enforcement happens in the policy layer, not in the antenna.
The second mistake is scoping the pilot too broadly. A pilot that covers all 60 AGVs and all 180 sensors from day one is a deployment with a pilot label. Start with one AGV travel corridor, one gNodeB, and a single GBR bearer for the e-stop channel. Prove that the QoS holds under cell load before expanding. The RF propagation behavior in your specific facility will surprise you — plan for it.
The technology should match the requirement. The requirement comes from your device inventory and your operations risk profile, not from a vendor's coverage map.