What Will Wi-Fi 8 Be? A Primer on IEEE 802.11bn Ultra High Reliability
The clearest short introduction: why reliability became the goal, and how multi-AP coordination builds on Wi-Fi 7's multi-link operation to get there.
Special report · IEEE 802.11bn Ultra High Reliability
Wi-Fi 8: wireless that behaves like a cable.
For twenty-five years every Wi-Fi generation sold the same promise: a bigger number on the box. Wi-Fi 8 is the first that doesn't — its peak speed is the same as Wi-Fi 7's. Instead, IEEE 802.11bn, code-named Ultra High Reliability, spends an entire generation on the moments you actually notice: the weak signal in the back room, the one late packet that stutters your game, the call that glitches as you walk between access points. The goal is wireless you stop thinking about, the way you never think about a network cable.
Wi-Fi 8 is still a draft standard — Draft 2.1 as of October 2026, with IEEE approval projected for May 2028. Features below follow the current 802.11bn draft and may still change. The diagrams animate; if your system is set to reduce motion, they hold a still frame.
Same average, very different worst case. Wi-Fi 8 is aimed at the spikes.
First, some history
Why Wi-Fi 6 took eight years
Wi-Fi 4 to Wi-Fi 5 took four years. Wi-Fi 5 to Wi-Fi 6 took eight — from the 802.11ac standard in December 2013 to 802.11ax in 2021. Since then a new generation has landed every three to four years. The long pause wasn't the industry going quiet. It was Wi-Fi rebuilding its foundations, and Wi-Fi 8 stands entirely on that rebuild.
The gap is where Wi-Fi stopped thinking like a shared Ethernet hub and started thinking like a cell network.
Ethernet thinking, then cellular thinking
Up to Wi-Fi 5, Wi-Fi worked like an old shared Ethernet hub: one medium, everyone listens, whoever finds it quiet gets to talk, and collisions get retried. Each generation made that one talker faster — wider channels, more antennas, denser modulation — but never changed the rules of who talks when.
Wi-Fi 6 changed the rules. The access point became a scheduler that hands out slices of time and frequency, the way an LTE or 5G base station does. That isn't an upgrade to the old design. It is a different design, and it had to be built from the radio up.
The same radio symbol from 1999 to 2013
802.11a defined Wi-Fi's OFDM symbol in 1999: 64 subcarriers per 20 MHz, 312.5 kHz apart, 3.2 µs long. Wi-Fi 4 and Wi-Fi 5 kept it exactly. They used more channels, more antennas and more bits per subcarrier — but the symbol itself never changed, so each new chip was an extension of the last.
802.11ax cut the same 20 MHz four times finer: 256 subcarriers, 78.125 kHz apart, a 12.8 µs symbol. That finer grid is what lets an access point carve a channel into resource units for several devices at once. It also meant a new preamble, new guard intervals, new pilots and new silicon — a genuine PHY reset.
Every comb line is a subcarrier. Four times as many, four times closer: room to share.
Uplink had to become a cellular uplink
Scheduling many devices to transmit to the access point in the same instant only works if each one lines up in time, frequency and power, on the AP's trigger. LTE handsets were built for that from day one. Wi-Fi clients never had been.
The yardstick changed
802.11ac was judged on peak speed: at least 1 Gbps. 802.11ax was asked for four times the average throughput per device in dense deployments. Before judging a single proposal, the group had to agree what "dense" looks like and calibrate simulators to measure it.
It failed ballot. Twice.
Draft 1.0 fell short of the 75% working-group approval it needed in early 2017. Draft 2.0 failed again that November, drawing 3,374 comments. Four more ballots and three sponsor-ballot rounds later, 802.11ax was approved in February 2021.
Shops filled the gap
Vendors kept selling 802.11ac, from 2016 as "Wave 2" with MU-MIMO and 160 MHz. The Wi-Fi Alliance renamed the generations in 2018 and certified Wi-Fi 6 against a draft in September 2019, so devices shipped well before the standard was final.
The reset was slow once. Everything built on top of it has been faster.
Wi-Fi 8 inherits the foundation, so it moves faster
Once the scheduled, OFDMA-based PHY existed, the next generations could build on it instead of replacing it. Wi-Fi 7 widened and bonded it. Wi-Fi 8 makes it dependable. That is why 802.11bn is projected to go from project approval to finished standard in under five years — the quickest of the four.
That is the shape of the whole story: Wi-Fi 6 was the PHY restart, Wi-Fi 7 scaled it, and Wi-Fi 8 is where the cellular ideas — scheduling, coordination, make-before-break mobility — are pointed squarely at reliability.
Durations from the IEEE 802.11 working group's official timelines. 802.11bn is a projection.
The brief
Three numbers, all about the worst case
When the IEEE approved the 802.11bn project it set three targets, each measured against Wi-Fi 7. None of them is a peak speed. All of them are about the bad moments.
25%
more throughput at the same signal quality — the weak, noisy link
25%
lower latency at the 95th percentile — the slowest 1 in 20 packets
25%
fewer lost frames — especially while roaming between access points
±0
added to peak speed — the first generation that isn't faster
The peak barely moves. What Wi-Fi 8 is graded on is the tail on the right.
Why the 95th percentile, not the average
Your drive to work averages twenty minutes. The average isn't the problem — the morning you sat behind a stalled lorry for forty-five minutes is. Latency works the same way. If 95 of 100 packets arrive on time, the average looks superb, and the other five are the stutter in your match, the frozen frame in the meeting, the robot arm that pauses.
So 802.11bn is graded on the tail. A 25% cut to 95th-percentile latency is something people feel immediately, even when the average barely moves.
Same peak as Wi-Fi 7 — deliberately
Wi-Fi 8 keeps Wi-Fi 7's ceiling: 2.4, 5 and 6 GHz, 320 MHz channels, 4096-QAM, up to eight spatial streams — about 23 Gbps on paper. (The 46 Gbps you may see quoted for both assumes sixteen streams, a configuration no consumer device uses.) Nobody has ever seen either figure in a home. The effort went where it pays.
Goal 1 — throughput where the signal is weak
Hearing the doorbell camera at the gate
A router has mains power, big antennas and room for heat sinks. The phone in the garden or the camera at the gate has a battery and a tiny antenna. So the link usually fails in the uplink first: the access point can shout, but it can't hear the reply. Wi-Fi 8's first two PHY features fix the reply.
A normal frame from the edge fades into the noise. The ELR frame, repeated, makes it.
Enhanced Long Range — slower, louder, repeated
When conditions get rough, the access point and the device agree to talk slowly and carefully — like ordering a drink in a noisy bar. An ELR frame uses a fixed 20 MHz channel, a single spatial stream and the two most robust modulations, BPSK or QPSK, for 1.67 or 3.33 Mbps. Its dedicated preamble has its training fields boosted by 3 dB so the receiver can find the frame at all.
The data itself is repeated across resource units: four copies of a 52-tone unit buy about 6 dB of extra margin. In the 5 and 6 GHz bands ELR is uplink-only — exactly where the link budget runs short. In 2.4 GHz it works in both directions.
Distributed resource units — every key at full volume
Think of a Wi-Fi channel as a piano keyboard, and the regulator as a rule about how loudly you may play any one octave. Normally a phone is handed a small cluster of neighbouring keys, so it has to play all of them softly to stay under the limit for that octave.
A DRU takes the same number of keys and spreads them across a 20, 40 or 80 MHz span. Each octave now holds only one or two of them, so each can be played at full volume. In the 6 GHz band, where indoor clients are capped at −1 dBm per MHz, that is up to about 11 dB more uplink power for the smallest units — without breaking a single rule.
Several devices can each use their own DRU at the same time, so the uplink OFDMA that Wi-Fi 6 introduced keeps working, just louder.
Same tones, same per-MHz limit. Spread out, every one of them can rise to the cap.
Finer rate steps
New intermediate modulation-and-coding steps fill the gaps between existing rates. When the signal dips, link adaptation drops half a rung instead of a whole one, so less throughput is thrown away.
Unequal modulation
With UEQM each spatial stream can use its own modulation. A strong stream carries more bits while a weak one stays robust, instead of every stream dropping to the level of the weakest.
Longer LDPC codewords
"2×LDPC" doubles the error-correcting codeword to 3,888 bits. A longer code has more redundancy to work with, which buys extra margin at the same data rate.
Goal 2 — the slowest 5% of packets
No more waiting at an empty junction
Most late packets aren't late because the air is full. They're late because the one slice of the channel everyone must check first is busy — often with a neighbour's traffic — while the rest of it sits idle.
Same busy primary. Wi-Fi 7 waits at the red light; Wi-Fi 8 drives round it.
Non-primary channel access — take the free lane
A wide Wi-Fi channel is a multi-lane road with one lane marked primary. Before sending anything, every device checks that lane. If it's busy, the whole connection sits at a red light — even with three empty lanes beside it. If your neighbour's network shares your primary lane, that happens all evening.
With NPCA, the access point announces a backup primary inside its own channel. When an overlapping network holds the main primary, devices switch to the backup and carry on, then switch back when the main lane clears.
Dynamic sub-band operation — one wide AP, several lanes at once
A Wi-Fi 8 access point may run a 320 MHz channel, but most phones and laptops top out at 80 or 160 MHz. Today they all crowd onto the same primary 80 MHz and take turns, while the rest of that expensive channel goes unused.
DSO lets the access point move each device onto a different slice of its wide channel — a short control exchange makes the switch — and then serve them in parallel. The wide channel finally earns its keep with everyday devices, not just the one flagship that supports it.
Three clients that can't use 320 MHz alone, using it together.
Prioritised channel access
Time-critical traffic gets a faster route to the air. A P-EDCA device can reserve the medium with a short defer-signal CTS instead of waiting out the full contention period — trimming the tail while the access point keeps the parameters tame enough that everyone else still gets a turn.
Low-latency indication
A device with something urgent queued can flag it inside a frame it is already sending. The access point schedules it straight away instead of discovering it a cycle later.
Coordinated wake windows
Wi-Fi 7's restricted TWT reserves protected service periods for latency-sensitive traffic. Wi-Fi 8's Co-RTWT lets neighbouring access points line those windows up so they don't land on top of each other.
Goal 3 — no lost packets on the move
Walk between access points without a hiccup
Every Wi-Fi generation so far, Wi-Fi 7 included, roams by tearing down one association and building another. Sometimes you don't notice. Sometimes the call drops a word.
Follow the playhead through the hand-off. Wi-Fi 7 drops packets in the gap; Wi-Fi 8 never opens one.
Seamless Mobility Domain — make before break
A Seamless Mobility Domain groups a set of access points under one management entity. To your device the whole group looks like a single network with several radios: it associates and authenticates once, with the domain, not with each access point.
- Prepare
- The target access point receives your context in advance — security keys, sequence numbers, Block Ack state, QoS settings — and sets up a dormant link.
- Execute
- The new link is switched on with no fresh association and no re-authentication. Multi-link operation lets it carry traffic before the old link stops.
- Complete
- Only then does the old access point release its state.
No teardown means no gap. It is the headline tool for the target of 25% fewer lost frames, and it needs Wi-Fi 8 at both ends — the access points and the device.
Access points that cooperate
From shouting match to traffic officer
Mesh systems already cooperate a little, but when it comes to actually sending data, every access point is on its own. In a mesh home or a block of flats that means radios on the same channel shouting over each other. 802.11bn adds Multi-AP Coordination: a standard way for access points to discover each other, agree a scheme, and then share the air on purpose. The one that wins the channel shares its transmit opportunity with a neighbour.
Co-TDMA — take turns, cleanly
The access point that wins the channel hands part of its transmit opportunity to its neighbour. No contention, no collision, no backoff — one schedule instead of two competing ones.
Co-SR — talk together, more quietly
Both access points transmit at the same instant, each turning its power down just enough that it doesn't drown the other's clients. Spatial reuse, negotiated instead of guessed.
Co-BF — aim away from each other
Both transmit at once, each steering a beam at its own client and a null towards the other's. It needs the two radios to share channel measurements and stay frequency-locked, which the draft defines.
In the current draft a coordinated transmission involves two access points at a time, set up with standard frames exchanged between them. A fifth scheme, Co-CR, lets neighbours agree channels for direct device-to-device links.
Your older devices benefit too
When a feature changes how a phone transmits or roams — ELR, DRU, seamless roaming — the phone has to support Wi-Fi 8. But coordination happens between the access points. A Wi-Fi 6 laptop on a coordinated Wi-Fi 8 network still lives on a calmer channel with fewer collisions. A Wi-Fi 8 router won't upgrade your old phone; it gives it a better-run network to live in.
Access points that can rest
Dynamic power save lets an access point — including a phone acting as a hotspot — idle in a low-capability mode and jump to full bandwidth and streams the moment a peer signals it has traffic.
Living with Bluetooth
Phones juggle Wi-Fi, Bluetooth and UWB on shared antennas. Wi-Fi 8 lets a device tell the access point when it will be unavailable, so frames aren't fired at a radio that's busy with your earbuds.
Better device-to-device links
Screen casting and PC-to-headset VR run direct between devices. The 802.11bn brief includes better peer-to-peer operation, so those links stop fighting the access point's own traffic.
Putting it together
What “feels like a cable” actually means
A cable doesn't feel solid because of its peak rate. It feels solid because it is predictable: the same latency every time, effectively no loss, nothing to re-negotiate when you move your chair. Each of those properties now has a matching Wi-Fi 8 feature.
| What a cable gives you | What breaks it on Wi-Fi today | What Wi-Fi 8 adds |
|---|---|---|
| The same link at any distance | Edge devices fall to the slowest rates, and the uplink fails first. | ELR, distributed resource units, finer rate steps, unequal modulation |
| Latency you can predict | One busy primary channel stalls everything; the 95th percentile spikes. | NPCA, dynamic sub-band operation, prioritised access, low-latency indication, Co-RTWT |
| No loss when you move | Roaming tears down one association before building the next. | Seamless Mobility Domain, make-before-break |
| Your neighbour's cable never touches yours | Overlapping networks contend, collide and back off. | Co-TDMA, Co-SR, Co-BF |
| No battery tax | Radios stay at full capability, waiting for traffic. | Dynamic power save, multi-link power management |
Deterministic, honestly
No radio can promise what copper does: a concrete wall still attenuates and a microwave still radiates. "Deterministic" Wi-Fi means the bad moments get rarer and shorter, and the network's behaviour becomes something you can plan around — latency and loss bounded with high probability, the way LTE and 5G engineer their service. That is the direction 802.11 has been travelling since Wi-Fi 6, and Wi-Fi 8 is the first generation to put it on the label.
How we measure it: real phones, real density
Tail latency and frame loss only show up under load, so a single laptop running a speed test can't reveal them. We test the way a mobile operator would: a phone farm of real iPhones and Android handsets, each one an independent client with real operating-system behaviour — idle chatter, bursts, background sync and roaming.
- The rig
- 2U racks holding 20 phones each. Sixteen of them around one access point puts 320 clients on a single AP.
- What it reveals
- Scheduling, airtime fairness and latency behaviour that single-client tests hide.
- What we see
- Wi-Fi 6/7 access points behave like small cells. BSS Coloring and OFDMA stabilise dense deployments. Performance scales with access-point density, not transmit power.
Density like a cell site, behaviour like real users. That is where the 95th percentile lives.
The router your ISP gave you
ISP equipment stops being the weak link
Most homes and shops never choose their Wi-Fi. It arrives with the fibre: one box, built to a price, mounted wherever the fibre comes in. Wi-Fi 8 doesn't make that box expensive. It moves the things that box gets wrong into the standard itself, so even an entry-level certified gateway gets the tools.
Same four boxes, same channel. The difference is whether they talk to each other.
Why ISP Wi-Fi disappoints today
- It lives in the wrong place. The box sits where the fibre terminates — a corner, behind the TV, inside a cabinet — and the far rooms get whatever leaks out.
- Every flat has the same one. A block where every unit got the same model on the same default channel is a block of radios contending with each other all evening.
- “Mesh” that doesn't coordinate. Satellites relay over the air on the same channel, and moving between them is a drop and a reconnect.
- Tuned for the speed test. One device next to the router looks superb. It says nothing about the 95th percentile with twenty devices in use.
What Wi-Fi 8 changes for an ISP box
- Bad placement
- ELR and DRU let distant phones and cameras still be heard on the uplink — the half of the link a corner-mounted router loses first.
- Identical neighbours
- NPCA routes round a neighbour sitting on your primary channel. Multi-AP coordination lets nearby access points take turns or share the air at lower power instead of colliding.
- Satellites & FTTR units
- A gateway and its own satellites are the easiest case for coordination: one operator, one management system. Add a Seamless Mobility Domain and walking room to room stops being a reconnect.
- Standard, not secret sauce
- Coordination is defined as standard frames between access points, so gear from different makers can cooperate rather than each mesh brand speaking only to itself.
- Graded on the tail
- The standard itself is judged on tail latency and loss, so the chipsets that end up in ISP gateways are built to that brief, not to a peak-speed screenshot.
What Wi-Fi 8 won't fix
- The bill of materials. The ISP still picks the chipset tier, the antenna count and the firmware. A Wi-Fi 8 logo on a two-antenna box is still a two-antenna box.
- Optional features. Much of 802.11bn is optional. Check what a gateway actually enables — coordination and seamless roaming especially — not just the generation on the label.
- Timing. Operators usually adopt a generation after certification, once chipsets reach their price point. Expect Wi-Fi 8 gateways in volume from 2028 onward.
- Physics. A wire still beats the best radio. Put access points on Ethernet, or on FTTR where cable can't go, and let the ISP box be the router it is good at being.
Stuck with an ISP router today? We'll tell you whether to tune it, wire around it or replace it
Where Wi-Fi 8 is heading
From draft to your ceiling — and on toward 2030
Dates come from the IEEE 802.11 working group's published schedule. Anything after October 2026 is a projection, and projections slip: Draft 1.0 and Draft 2.0 each landed two months late.
-
The project is approved
The 802.11 working group approves the Ultra High Reliability project request in July; the IEEE-SA Standards Board approves P802.11bn in September. The scope: 25% better throughput, 95th-percentile latency and frame loss than Wi-Fi 7.
-
Task Group bn meets
TGbn holds its first meeting and starts sorting hundreds of technical proposals into a feature set.
-
Draft 0.1
The first consolidated draft text of the amendment.
-
Draft 1.0 goes to ballot
The first working-group letter ballot, two months behind plan. Thousands of comments follow, as with every Wi-Fi generation.
-
First silicon
MediaTek (Filogic 8000) and Broadcom show Wi-Fi 8 chips at CES 2026; Qualcomm announces FastConnect 8800 at MWC in March. All of it built to a draft, ahead of certification.
-
You are here · October 2026
Draft 2.0, then 2.1
Draft 2.0 goes to letter ballot in July. At the September interim the group resolves about 1,850 comments and produces Draft 2.1.
-
Draft 3.0 ballot
The next working-group letter ballot. Projected.
-
Draft 4.0 to IEEE-SA ballot
The wider standards-association ballot begins. Projected.
-
Wi-Fi CERTIFIED 8
The Wi-Fi Alliance is expected to launch certification around CES 2028 — the point where interoperable products ship in volume.
-
802.11bn becomes a standard
Final working-group and 802 Executive Committee approval in March, IEEE-SA approval in May. Projected.
-
Into offices, homes and ISP gateways
Enterprise access points and flagship phones first, then mainstream laptops and operator gateways. By 2030, Wi-Fi is designed and measured on reliability, not on the number on the box.
Beyond Wi-Fi 8
802.11bn isn't the only thing in the pipeline. These run alongside it in IEEE 802.11 today.
The direction of travel
Protect the new. Out-reuse the old.
Since Wi-Fi 6 the industry's stance has quietly changed: all generations must coexist politely forever has given way to modern Wi-Fi must work, even if legacy devices suffer. Wi-Fi 6 and later access points can raise their OBSS-PD threshold from −82 dBm up to −62 dBm — ignoring weak overlapping transmissions and carrying on at slightly lower power.
Devices from 802.11a/b/g/n/ac can't take part in the scheduling, so more and more they are treated as background energy rather than peers to be waited for. The practical upshot: retire Wi-Fi 5-and-older clients where they matter, and keep the stragglers on a band of their own.
From Wi-Fi 5's perspective“Why is my throughput suddenly terrible?”
From the standard's perspective“You're just unmanaged background energy now.”
Read the research
IEEE papers and primary sources
Everything on this page traces back to these. The IEEE magazine articles are the best place to start, the tutorials go deeper, and the working-group pages are where the schedule actually lives.
Future Directions for Wi-Fi 8 and Beyond
Written before 802.11bn existed: the market drivers and candidate directions — reliability, latency, new bands — that became the UHR brief, with Wi-Fi 8 forecast for 2027–2028.
A Tutorial on IEEE 802.11bn Multi-AP Coordination for Wi-Fi 8: From Standardization to Performance Evaluation
The deep dive on Co-TDMA, Co-SR, Co-BF and Co-RTWT, with an open-source Wi-Fi 8 simulator used to test whether coordination can hit the 25% targets.
A Tutorial on Wi-Fi 8: The Journey to Ultra High Reliability
Feature-by-feature: distributed-tone resource units, unequal modulation, prioritised and non-primary channel access, roaming and multi-AP coordination.
Towards Ultra-High Reliability in Wi-Fi 8: IEEE 802.11bn Core Mechanisms, mmWave Integration, and Performance Verification
System-level simulations of whether the 802.11bn mechanisms meet their targets, plus how integrated mmWave could extend Wi-Fi's spectrum.
Wi-Fi: Twenty-Five Years and Counting
From 802.11b to 802.11bn, organised by mechanism rather than generation — the best long read on why each generation looks the way it does.
A Tutorial on IEEE 802.11ax High Efficiency WLANs
The Wi-Fi 6 PHY reset in detail: OFDMA, the new numerology, trigger-based uplink — the foundation everything on this page sits on.
IEEE 802.11ax: High-Efficiency WLANs
Written while 802.11ax was still being designed: why dense deployments forced a different approach, and part of the answer to why it took so long.
Standards and schedules
- IEEE P802.11 TGbn status
- The task group's own page: scope, objectives, draft milestones and the projected approval date.
- IEEE 802.11 official timelines
- Ballot and approval dates for every amendment — the source for the project-duration chart above.
- IEEE 802.11 working group
- Current task and study groups, including 802.11bq, bp, bt, bi and AI Offload.
- Samsung Research — IEEE 802.11bn UHR
- A clear engineering walkthrough from a TGbn contributor; the source for the ELR and DRU figures.
Wi-Fi 8 is a draft. Feature names, parameters and dates on this page follow the IEEE 802.11bn draft and working-group reports as of October 2026, and may change before the standard is approved. Some everyday analogies here — the noisy bar, the piano keyboard, the highway — follow Techquickie's "What is Wi-Fi 8?" explainer.
Questions
Wi-Fi 8, answered
Only if you're buying in 2028 anyway. Certified Wi-Fi 8 products are expected around the start of 2028, and the features that matter most need Wi-Fi 8 phones and laptops too. If your network struggles now, a properly wired Wi-Fi 7 or 6E deployment fixes most of it today — and Wi-Fi 8 is fully backward compatible when you do move.
Not at the peak; the ceiling is identical. It is meant to be faster where it counts — at the edge of coverage, in a crowded channel and in the slowest moments — which are the speeds you actually experience.
Yes. They connect normally and benefit indirectly from coordination between access points. Client-side features such as ELR, distributed resource units and seamless roaming need Wi-Fi 8 in the device as well.
Chips and a few early routers are appearing, built to a draft. Like early Wi-Fi 6 and 7 gear, they may need firmware updates as the standard settles, and some features may arrive later or not at all. For a business network we would wait for certification.
Closer than any Wi-Fi before it, and measured on the right things. But a radio is still a radio. For anything that truly can't drop — a POS terminal, a NAS, a workstation that never moves — use the cable. Use Wi-Fi 8 for everything that moves.
Eventually. Operators typically adopt a generation once it's certified and the chipsets reach their price point, so expect Wi-Fi 8 gateways to roll out from 2028. Until then, the biggest win is usually adding wired access points behind the ISP box — see FTTR for buildings that can't take new cable.
Planning a network that should still be right in 2030?
We design for the generation you can buy today and the one coming next: wired backhaul, sensible access-point density and hardware with a clear upgrade path. No guesswork, no upselling to the newest number.