UniFi Wi-Fi 7 and MLO: When 6 GHz Is Worth the Upgrade, and When It Isn't

Do you know how many devices on your network can actually associate on 6 GHz right now? If you are about to sign off on a fleet of U7 access points, that number decides whether the upgrade earns anything at all.

Wi-Fi 7 has quietly become the default line item in the UniFi catalogue. The U7 Pro, U7 Pro Max, U7 Pro Wall, U7 Enterprise, and E7 have displaced the U6 series in most quotes, which means a lot of operators are buying 802.11be whether or not they set out to.

That said, the three headline features — 320 MHz channels, 4096-QAM, and Multi-Link Operation — each carry a precondition that has nothing to do with the access point itself. Two of those preconditions live in your client fleet, and the third lives in the switch closet.

What follows is the calculus: what 802.11be actually changes on the air, what MLO requires before it will turn on, why 6 GHz costs you coverage, and which sites should stay on Wi-Fi 6 or 6E for another refresh cycle.

Wi-Fi 7 is an amendment with one genuinely new architectural idea and three efficiency improvements stacked on top of 802.11ax. Knowing which is which keeps you from paying for capability your clients cannot reach.

Here is what the standard adds, and what each addition depends on:

• 320 MHz channels. Double the 160 MHz maximum of Wi-Fi 6E, and available only in the 6 GHz band. The US allocation from 5.925 to 7.125 GHz yields three non-overlapping 320 MHz channels, which is your entire reuse plan in any building with more than three access points in earshot of one another.
• 4096-QAM. Encodes twelve bits per symbol where 1024-QAM manages ten, lifting the very top of the rate table. It demands an extremely high signal-to-noise ratio, so in practice you see it in the same room as the access point and essentially nowhere else.
• Multi-Link Operation. A single association that spans two or three radio links at once instead of pinning a client to one band for the life of the session. This is the only Wi-Fi 7 feature that changes the shape of the connection rather than the speed of it.
• Multi-RU and preamble puncturing. An access point can keep a wide channel in service with an interfered or radar-occupied subchannel punched out, instead of collapsing to the next width down. This is the feature most likely to help a messy real-world site, and it asks nothing at all of your users.

Notice that three of those four only pay out at close range, on clean spectrum, to a client radio built in the last two years. The fourth pays out on bad sites, which is where most deployments actually live.

Keep in mind that none of it raises the ceiling imposed by the wired side. An 802.11be radio negotiating a headline PHY rate over a 1GbE uplink is still a gigabit access point, and the client status page will happily show you the higher number anyway.

MLO lets a client hold one logical association while maintaining links on 5 GHz and 6 GHz — and in some implementations 2.4 GHz — at the same time. Traffic moves between those links without a reassociation, a new DHCP lease, or a fresh key exchange.

There are two implementation classes and they buy you very different things. STR, or simultaneous transmit and receive, genuinely runs both links at once and can aggregate across them; eMLSR keeps both links alive for listening while transmitting on only one at a time.

Most of the client fleet you will meet in the field is eMLSR rather than STR, including a large share of the laptop and handset silicon shipping under the Wi-Fi 7 badge. The honest benefit there is faster band selection and a shorter stall when one link goes busy, which shows up as a better P95 latency figure and not as a bigger speed test result.

MLO also carries a hard prerequisite that catches people mid-rollout. It is defined only over WPA3-SAE with protected management frames required, so there is no MLO on WPA2 — and a WPA2/WPA3 transition-mode SSID will not carry it either.

Be aware of what that means for SSID design. Turning MLO on for your one flat corporate SSID strands every WPA2-only printer, badge reader, sensor, and older handset that cannot complete an SAE handshake.

The workable pattern is a dedicated WPA3-only SSID carrying MLO and 6 GHz for modern clients, with legacy device classes left on their own SSID and their own VLAN. If you have not already separated those classes, UniFi VLAN segmentation is the prerequisite project rather than an optional cleanup afterwards.

MLO manages links to the access point a client is already associated with, and nothing more. Moving between access points is still 802.11k, 802.11v, and 802.11r plus a sane minimum RSSI, so your UniFi roaming configuration matters exactly as much as it did on Wi-Fi 6.

Free-space path loss rises with frequency, so a 6 GHz signal arrives weaker than a 5 GHz signal sent at the same power across the same distance. Add drywall, glass, concrete, and metal studs, and that gap widens rather than holding constant.

Regulatory limits push in the same direction. Indoor 6 GHz operation in the United States runs under Low Power Indoor rules unless the radio is coordinated through AFC, and LPI caps EIRP while prohibiting external antennas — which describes essentially every indoor UniFi model you are quoting.

Channel width compounds the problem. Doubling a channel raises the noise floor by roughly 3 dB, so a 320 MHz channel needs materially more signal to sustain the same modulation than an 80 MHz channel does.

That is why the client showing 320 MHz at your desk shows 160 or 80 two rooms away, assuming it stays on 6 GHz at all. Band steering decisions that looked clean in the controller get made again, badly, by client firmware you do not control.

Discovery behaves differently up there as well. There is no legacy probe process in 6 GHz — clients find the band out of band, through Reduced Neighbor Report elements carried in the 2.4 and 5 GHz beacons and through FILS discovery frames, so a 6 GHz-only SSID with no lower-band counterpart is findable but slow and inconsistent across vendors.

All of this adds up to one design rule. Treat 6 GHz as a capacity layer laid over a 5 GHz coverage layer, and size the deployment on 5 GHz plus the 2.4 GHz tail of scanners, sensors, and handhelds.

Swapping U6 hardware for U7 hardware in the same mounting positions gives your high-value clients somewhere clean to go. It does not push usable signal into corners that were already marginal, which remains an access point count and placement problem no amendment will fix for you.

The access point is the cheap part of a Wi-Fi 7 refresh. Everything it plugs into is the expensive part, and that is where quotes tend to go quiet.

Start with the uplink. A Wi-Fi 7 access point on a 1GbE port is capped at a gigabit of wired throughput regardless of what the radios negotiate, which reduces 320 MHz and 4096-QAM to decoration.

U7 Pro and U7 Pro Max ship with 2.5GbE uplinks, and the higher-tier U7 and E7 models step up to 10G. None of that matters if the far end of the run terminates on a 1G access port, so a real Wi-Fi 7 rollout is a UniFi switch upgrade with new radios attached — budget it that way from the first draft.

Power is the second line item. U7-class access points generally require 802.3at PoE+ rather than 802.3af, and the 10G models want 802.3bt PoE++, so confirm the class on the datasheet for the exact SKU because it varies inside the family.

Multiply that across a floor and the switch's total PoE budget, not its port count, becomes the binding constraint. Run that arithmetic before you order — our PoE budget planning guide covers the headroom and derating rules that catch people on high-density floors.

The cable plant is rarely the blocker. Both 2.5GBASE-T and 802.3bt run happily over existing Cat5e at normal in-building lengths, which is precisely why the money lands on the switch rather than the conduit.

Finally, be honest about where the traffic is going. If most of what your users do is internet-bound over a handoff measured in a few hundred megabits, the air was never the constraint, and edge capacity and resilience — including WAN failover on the gateway — will change the experience more than a new radio does.

Wi-Fi 7 helps with LAN-side transfers, dense-client contention, and latency under load. It does not manufacture WAN bandwidth, and no channel width will.

There are deployment profiles where Wi-Fi 7 clearly earns its cost. They share one trait: the air is genuinely the bottleneck, and the client fleet is modern enough to exploit the fix.

• High-density spaces with modern client fleets. Lecture halls, conference centres, trading floors, and clinics where 5 GHz is already congested and a meaningful share of devices are 6 GHz capable. Contention relief, not peak rate, is what you are buying.
• Greenfield builds and full cabling refreshes. Specifying 2.5GbE and PoE+ switching during construction costs very little at the margin, while retrofitting the same capability into a live building costs a great deal.
• Sites already running 6 GHz on Wi-Fi 6E with real client counts on it. You have already proved the fleet will use the band, which means MLO and wider channels have somewhere to land on day one.
• Interference-heavy environments. Multi-RU puncturing keeps wide channels in service through DFS radar hits and noisy neighbours that would otherwise force a width drop for every client on the radio.
• Latency-sensitive wireless workloads. Wireless AV, voice handsets, handheld point-of-sale, and telemetry all benefit from MLO's link redundancy even when raw throughput is beside the point.

And here are the profiles where a U7 refresh spends money without returning much of it:

• Fleets dominated by 2.4 GHz IoT and Wi-Fi 5 clients. Warehouse scanners, badge readers, thermostats, and older VoIP handsets receive exactly the same service from a U6 as from a U7. Their constraint is airtime discipline and 2.4 GHz channel planning.
• 1GbE switching you have no plan to replace. You would be paying for radios the uplink cannot drain, and the client status page will report rates the network never delivers.
• Coverage complaints. Dead spots are an access point count and placement problem, and a newer standard mounted in the same positions changes nothing about the far corner of the building.
• WPA2-locked device classes. Medical carts, industrial controllers, and some building-management gear cannot complete SAE at all, which puts MLO permanently out of reach for the SSID they live on.
• Sites whose real constraint sits upstream. When the WAN circuit or the uplink to the core is saturated, fix that first and revisit the radios afterwards.

If your environment sits between those two lists, phase it rather than deciding it. Put U7 hardware in the two or three highest-density rooms that already have 2.5GbE ports available, stand up a WPA3-only SSID with MLO enabled alongside your existing network, and leave both running for a month.

The data you want from that month is straightforward: how many clients associate on 6 GHz, what channel width they actually negotiate, and whether measured 5 GHz load drops as a result. Anything short of that is a catalogue decision wearing an engineering costume.

If you are scoping the refresh now, start with the client audit, then match hardware to what it tells you — our UniFi access point comparison and UniFi switch comparison pages lay out uplink speed, PoE class, and radio configuration side by side so the switch and the access point get specified together.