For twenty years, every generation of Wi-Fi arrived with the same argument: a bigger number than the one before. Wi-Fi 8, which reaches shelves over the next few months, breaks the tradition — the theoretical peak stays where Wi-Fi 7 left it, around 23 Gbps — and moves the promise onto different ground: making the connection work well when the signal is weak, the house is full of devices, and you are in the wrong room.
What it actually promises
The technical name is IEEE 802.11bn, and the acronym the working group uses says it all: UHR, Ultra High Reliability. The stated goals are three, and they are all improvement percentages rather than absolute speeds:
- around 25% more throughput in weak-signal conditions, meaning far from the router;
- around 25% lower latency measured at the 95th percentile, which is the technical way of saying “the worst moments get less bad”;
- around 25% fewer dropped packets, with fewer retransmissions and fewer micro-interruptions.
It is a shift in priorities that makes sense: in an ordinary home, peak speed has never been the bottleneck. The problem is the video call that stutters when someone turns on the microwave, the game that loses a second at the wrong moment, the file that copies at 40 MB/s in the attic instead of 400 in the living room.
Why “reliable” is harder than “fast”
Raising peak speed is, roughly, a matter of wider channels and denser modulation: Wi-Fi 7 reached 320 MHz and 4K-QAM, and there is not much headroom left. Reliability, on the other hand, is won through coordination: access points that talk to each other instead of fighting over the same channel, transmissions scheduled together, devices that can lean on more than one access point at once.
It is the same reasoning that led Wi-Fi 7 to introduce Multi-Link Operation, the ability to use several bands together on one connection. Wi-Fi 8 takes that road further, towards networks where several devices behave as a single system rather than as noisy neighbours.
Do you need a new router?
Almost certainly not, not now. The first chips appeared at CES in January, the first consumer routers are announced for October, but the standard will not be finalised before 2028: what you buy today is based on a draft, and full compliance will depend on firmware updates nobody can guarantee in advance.
Meanwhile, Wi-Fi 7 is at its best: the hardware is mature, prices have come down, and the devices that support it now number in the hundreds of millions. If your home network struggles, the protocol generation is rarely the reason.
The open problems
- Your devices decide. A new router is useless if your phone, laptop, and TV do not speak the same generation. Installed hardware turns over in five or six years, not one.
- The 6 GHz band does not go through walls. It is wide and uncrowded, but concrete stops it. In many homes the real jump comes from a cable or a second access point, not a new protocol.
- Buying on a draft. Anyone who lived through the “draft” phases of previous generations knows the pattern: it works, with incompatibilities you discover by using it.
- The line upstream. If your fibre delivers 1 Gbps, no Wi-Fi will make the internet faster. This improves what happens inside the house, not what arrives from outside.
Tip: before shopping for routers, measure. Run a speed test over a cable and one over Wi-Fi in the room where the connection is worst. If the cable is fine and Wi-Fi is not, the problem is coverage, and it is solved by where the device sits or by a second node. If the cable is bad too, the router is not the culprit.
What to expect next
The realistic calendar is this: draft hardware between late 2026 and 2027, Wi-Fi Alliance certification in 2028, real adoption in phones and laptops from 2029. Anyone buying a router today should choose it for what it can do now — 6 GHz, mesh with a dedicated backhaul, decent handling of legacy devices — and not for the number printed on the box.
The most interesting thing about Wi-Fi 8 is not technical but cultural: it is the first generation to openly declare that the problem to solve was never speed. It only took twenty years.