Every new generation of Wi-Fi arrives carrying higher data rates than the last. Wi-Fi 5 hit 1 Gb/s, Wi-Fi 6 pushed into the high single digits, and Wi-Fi 7 now advertises a peak near 46 Gb/s. That’s more than an order of magnitude in a decade. But a data rate is only as useful as the devices that can act on it, and most of them never will. Few people replace their phones, laptops, and routers in step with each release of the standard, so the installed base trails the specification by years. Set that aside, and a more basic question remains, one the rising numbers rarely invite: how much throughput does anyone actually need?
Past roughly 10 Gb/s, the practical benefit of going faster begins to fade. It is hard to picture a home or office where many devices are contending for the air at once, where a user would notice the jump from 10 to 20 Gb/s. None of this makes the advances hollow, since each new version also brings real gains in efficiency, latency, and the handling of crowded environments. It is only that the headline figure, the one number every launch is built around, may be the part that matters least.
Start with that number. Reaching it requires 320-MHz channels, the highest modulation order, and sixteen spatial streams operating simultaneously. No client device implements sixteen streams. Phones implement two. Laptops implement two, occasionally three. The advertised rate describes a configuration that exists in no product a consumer can buy, decoded by a receiver that has never been built into anything you would carry. It is better understood as a theoretical upper bound than as a specification any user will meet.
The asymmetry is structural and worth stating in engineering terms. An access point can support eight or more antenna chains, mains power, and a thermal envelope that allows it to sustain the highest modulation continuously. A handset has two antennas if you are lucky, a battery to protect, and a front end built to a bill of materials. The advertised throughput scales linearly with spatial streams, so a two-stream client starts at one-eighth of the headline before any other impairment enters the picture. Everything after that only widens the gap.
Consider the modulation. Wi-Fi 7’s signature feature is 4096-QAM, which packs 12 bits per symbol and delivers roughly 20% more throughput than 1024-QAM. However, 4096-QAM requires a signal-to-noise ratio of 42 dB, versus 31 dB for 1024-QAM in Wi-Fi 6 and 25 dB for 256-QAM in 802.11ac. That is not a marginal increase: Run the budget for a 160 MHz channel, with a thermal noise floor near -92 dBm, and a typical receiver noise figure of 6 dB lifts the effective floor to about -86 dBm. To achieve a 38 dB SNR for 4096-QAM, the signal at the receiver must be roughly -48 dBm, with modest transmit power and antenna gain. That corresponds to a separation of about 1.5 m between client and access point. The top modulation order is a feature you can use while holding the phone against the router.
Walk into the next room, and the rate adaptation algorithm has already stepped you down the MCS table toward a figure well below the one shown in giant letters on the box it came in. Apple’s recent iPhones limit Wi-Fi 7 modulation to MCS 11 and do not implement 4096-QAM indices at all. One of the most widely sold premium handsets on the market omits the headline feature of the standard it supports because the silicon, power budget, and antenna just don’t justify it. A modulation order reachable only at arm’s length from the access point means little in a device meant to be carried, and the handset design reflects that, even if the specification sheet still leads with the larger number.
Now let’s take a look at the spectrum. The 320 MHz channels that the peak rate assumes exist only in the 6 GHz band, and that band’s availability and power rules vary by region, with much of the world having opened less of it than the U.S. With indoor and standard power operation governed by different limits and, in the standard-power case, by automated frequency coordination to protect incumbents. A wide channel also collects more noise, since doubling the bandwidth raises the integrated noise floor by 3 dB. That means the channel width that inflates the peak rate also raises the SNR bar the client must clear to use it. The two headline ingredients, the widest channel and the densest modulation, work against each other in any environment that is not a shielded room.
Even in ideal (i.e., never experienced) conditions, the air interface has not been the bottleneck for several generations. A client capable of delivering 2 Gb/s over the air is still bound by a 300 Mb/s access plan, congested backhaul, and a server on the far end that was never going to feed it that fast. For most of the real traffic, the Wi-Fi link had headroom to spare since Wi-Fi 5. Doubling a ceiling nobody touches produces a better spec sheet and an identical experience.
So why does every generation lead with a bigger number? Because the number is legible. Peak throughput fits on a box, sorts cleanly in a comparison table, and lets a buyer believe the more expensive router is the faster one. The genuinely valuable work, the scheduling and coordination that determine whether a video call survives a crowded conference room, does not reduce to a single digit. Latency, jitter, and determinism are what users actually feel, and none of them headline a press release. The industry has spent two decades leading with the metric that may matter least to the people buying it.
What makes this worth saying now is that the standards body has finally said it, too. Wi-Fi 8, the IEEE 802.11n amendment, declines to raise the peak rate. It supports the same 320 MHz channels, 4096-QAM, eight spatial streams, and peak PHY rate as Wi-Fi 7. For the first time, a new generation keeps the headline figure flat while putting its engineering elsewhere. The stated goals are at least 25 percent higher throughput in challenging signal conditions, a 25 percent reduction in 95th-percentile latency, and 25 percent fewer dropped packets when a device roams between access points, all under a banner the committee named Ultra High Reliability. The draft reached version 1.0 in 2025, with publication targeted for March 2028.
The mechanisms tell you where the real problems were hiding. Wi-Fi 8 introduces coordinated spatial reuse and coordinated beamforming so neighboring access points cooperate rather than interfere, dynamic sub-channel operation to use fragments of spectrum that would otherwise sit idle, and enhanced modulation and coding schemes aimed at the low end of the SNR range rather than the top. The most telling addition is at the physical layer.
Distributed resource units spread a transmission’s tones across a wider slice of bandwidth to raise the per-tone transmit power, and enhanced long-range frames use power-boosted preambles and frequency-domain duplication to extend uplink coverage. That feature exists because the client is power-limited and constrained by spectral-density rules, so the only way to improve its uplink is to use the spectrum mask more effectively rather than demanding a higher modulation it can never achieve. This is the opposite of the peak-rate philosophy. It is engineering for the device that actually exists.
Read the scope document, and the subtext is unmistakable. The committee responsible for the headline figure has concluded that it is no longer a source of meaningful gains. The remaining problems are reliability problems. The edge user with a weak signal, the warehouse with 200 contending clients, and the factory floor that needs a packet to arrive on time rather than quickly. None of these is solved by another doubling of a rate the client cannot reach. They are solved by coordinating access points, by extending uplink range within a power budget, by making the worst case less bad rather than the best case more theoretical.
There is a lesson here for anyone specifying a network, and it runs counter to the buying process. The right question was never how many gigabits the radio claims. It was how the link behaves under ordinary conditions, which is to say, most of the time. A system that delivers 200 reliable Mb/s to every device in a full room is worth more than one that promises 2 Gb/s to a single device a meter from the antenna, and the gap between those two systems is exactly the gap the peak rate conceals.
Wi-Fi 8 will ship with a smaller headline figure than its predecessor, but a more useful set of improvements. The open question is how to communicate that. A reliability gain shows up mainly by its absence, in the call that did not drop and the file that transferred without a stall. The engineers writing the standard have already made their choice about what matters. The rest of the industry now has to find a way to convey a benefit that resists being reduced to a single number.