Mobile generations are announced with three claims: much faster, much lower latency, and vastly more connected devices. 5G made all three. It delivered convincingly on one, partially on another, and the third turned out to matter mainly to industries most people never see.
None of that makes 5G a failure. It makes the gap between a specification and a deployment worth understanding, because the same gap is about to open again.
What 5G actually promised
The formal targets, as written into the standards process:
- Peak rates around 20 Gbps downlink
- Latency approaching 1 ms over the air
- A million connected devices per square kilometre
- 99.999 per cent reliability for critical applications
- Substantially better energy efficiency per bit
Every one of those numbers is achievable in a laboratory. The question was always which would survive contact with real spectrum, real geography and real capital expenditure.
What actually arrived
Capacity, which was the real win
The genuine achievement of 5G is not that any individual gets a faster connection. It is that a stadium of forty thousand people can all use their phones at once.
Anyone who attended a large event before roughly 2020 remembers the total network collapse. That has largely stopped. 5G's improvements in spectral efficiency and dense deployment mean crowded places work now.
This is a substantial and genuinely valuable improvement. It is also almost impossible to advertise, because the benefit is the absence of a problem.
Speed, conditionally
Real-world 5G speeds vary by more than an order of magnitude depending on which spectrum band you are actually connected to, and phones do not tell you clearly.
| Band | Typical real speed | Coverage from one site | What it is used for |
|---|---|---|---|
| Low (600-900 MHz) | 30-120 Mbps | Many kilometres | Rural and nationwide coverage claims |
| Mid (2.5-4.9 GHz) | 150-700 Mbps | 1-3 km | The band that carries actual traffic |
| High (24-40 GHz) | 1-3 Gbps | 200-500 metres | Stadiums, airports, dense city blocks |
Mid-band is where 5G lives and works. High-band, the millimetre wave spectrum that produced all the gigabit demonstrations, is blocked by walls, foliage, rain and hands, and has been deployed sparingly because the site density required is economically brutal.
If your 5G feels like slightly better 4G, you are probably on low-band. That is not a fault; it is the coverage trade-off working as designed.
Latency, partially
Over-the-air latency did improve substantially, from around 30-50 ms on 4G to roughly 8-20 ms in good 5G conditions.
The sub-millisecond figure requires edge computing — putting the server physically near the tower — and that has been deployed in limited locations. The radio was never the whole delay. If the server is 600 km away, the radio improvement is a rounding error, a point we cover in more detail in our piece on where cloud gaming latency comes from.
Network slicing, quietly
The feature that got least attention and delivered most reliably. Slicing lets an operator run logically separate networks on shared infrastructure with guaranteed characteristics — a slice for emergency services that stays available when a crowd saturates everything else, a slice for a factory floor with hard reliability guarantees.
This has real industrial value. It has no consumer story at all, which is why nobody has heard of it.
The parts of 5G that worked were the unglamorous ones: capacity, efficiency, industrial guarantees. The parts that were advertised — gigabit phones, remote surgery, self-driving fleets coordinating over the air — were either niche or dependent on things other than the network. Expect the same asymmetry from 6G.
What did not happen
Worth being direct about, because these were used to justify substantial public investment.
Remote surgery. Latency was never the obstacle. Liability, regulation, and the fact that surgeons prefer to be in the room were.
Vehicles coordinating over cellular. Autonomous systems are being built to work without connectivity, because a car that requires a network is a car that fails in a tunnel.
The internet of everything. Massive device connectivity is real and used, but mostly by long-standing low-power technologies that predate 5G and cost less to run.
Fixed wireless replacing fibre everywhere. It has become a genuinely good option in areas where fibre is uneconomic, which is a real success, but it is a complement rather than a replacement.
What 6G is likely to be
Standardisation work is under way, with commercial deployment realistically toward the end of this decade. The recurring themes:
Sensing as a first-class function. The most genuinely novel idea. Radio signals reflect off objects, so a network that is already transmitting everywhere can also perceive its environment — presence, motion, rough shape, even breathing rate. Useful for safety and automation, and obviously a significant privacy question that is not yet settled.
AI-native radio management. Using learned models rather than fixed algorithms for beam steering, interference management and resource allocation. Efficiency gains, invisible to users.
Higher frequencies still. Sub-terahertz bands, with enormous bandwidth and even worse propagation than millimetre wave. Expect very localised deployment.
Integrated satellite coverage. Direct-to-phone satellite links treated as part of the network rather than a separate service. This is the change most likely to matter to ordinary people, because it addresses coverage rather than speed.
Energy efficiency as a headline target. Networks are a meaningful electricity consumer and operators have strong commercial reasons to care.
A reasonable expectation
Based on how the last three generations went, a sober forecast for 6G:
- Peak speeds will be quoted in hundreds of gigabits and will not be experienced by consumers.
- Coverage will improve mainly because of satellite integration, and that will be the genuine win.
- Latency will improve modestly, and will still be dominated by where the server is.
- Sensing will find real uses in industry and will trigger a privacy debate that should start earlier than it will.
- The consumer benefit will again be capacity: more people using more data in the same places without degradation.
What this means for you
Practical implications, without the hype.
- Do not buy a phone for 6G support until networks exist. Early-generation modems are consistently worse than second-generation ones.
- Check band support, not generation labels, if you care about performance. Mid-band coverage is what determines your experience.
- Coverage maps beat speed claims. A network that works where you live at 100 Mbps is better than one that reaches 900 Mbps three streets away.
- For anything latency-sensitive, the server location matters more than the radio. No mobile generation will fix a badly placed data centre.
- Fixed wireless is worth checking if you have no fibre option. It has quietly become good.
The revolution language is not going to stop. But mobile generations have never been revolutions — they have been steady, expensive, largely invisible increases in capacity, punctuated by one or two features that turn out to matter for reasons nobody put on a billboard. 5G's was handling crowds. 6G's will probably be covering the places that currently have nothing.