5G and Beyond: What Changes When the Network Gets Smart

5G was sold to consumers as faster downloads, which was the least interesting thing about it. What 5G actually introduced was a network that could be programmed: carved into virtual slices with different guarantees, extended with compute at the tower, and scaled to millions of low-power devices per square kilometre. Several years into deployment, it's worth separating what materialised from what didn't — and what the 6G research programme is chasing.

Three networks in one

The 5G specification defines three service profiles, and they matter more than the headline speed:

• Enhanced mobile broadband (eMBB): the consumer story. Multi-gigabit peak rates, especially on millimetre-wave spectrum; in practice, mid-band (3–4 GHz) delivering a few hundred megabits with good coverage is what most people experience. Fixed wireless access — 5G as a home broadband replacement — turned out to be one of the biggest commercial wins.
• Ultra-reliable low-latency communication (URLLC): targets of 1 ms air-interface latency and 99.999% reliability for industrial control, remote operation and vehicle safety. Deployed mainly in private networks so far, not the public macro network.
• Massive machine-type communication (mMTC): up to a million devices per km² at very low power, for sensors that report a few bytes an hour. Delivered largely through NB-IoT and LTE-M, which 5G absorbed.

Network slicing

Because the 5G core is software running on commodity servers, an operator can create logically isolated networks — slices — over the same physical radio and transport, each with its own bandwidth, latency and reliability guarantees. A hospital's telemetry, a stadium's video streaming and a logistics fleet's tracking can share a tower while being scheduled and secured separately. Slicing works today, particularly in private 5G deployments at ports, mines, factories and campuses, where an enterprise owns the network end to end. On public networks it has rolled out more slowly, largely because the commercial model — who pays for a guaranteed slice, and how much — took longer to settle than the technology.

Compute at the tower

Low latency on the radio link is wasted if the application server is 800 km away. Multi-access edge computing (MEC) places servers inside the operator network, near the base stations, so that applications — AR rendering, vehicle coordination, video analytics — run a few milliseconds from the device. This is where 5G and edge computing become the same conversation, and where the hyperscalers have partnered with carriers to put cloud regions inside telecom facilities.

What was harder than expected

• Densification cost. High-band 5G needs small cells every few hundred metres. Deploying them — sites, power, fibre backhaul, permits — proved slow and expensive, which is why millimetre-wave coverage remains patchy and mid-band became the workhorse.
• The "killer app" question. Consumers didn't need more speed than 4G offered for most uses. The revenue growth came from enterprise: private networks, fixed wireless, IoT connectivity.
• Security surface. A software-defined, cloud-native core is more flexible and also more attackable. Supply-chain concerns about network equipment vendors became a geopolitical issue, and the shift to Open RAN (interoperable radio components from multiple vendors) was driven as much by that as by cost.
• Energy. 5G radios are more efficient per bit but there are far more of them; total network energy consumption rose, making power management a first-class engineering concern.

Standalone 5G and 5G-Advanced

Early "5G" mostly ran a new radio on top of a 4G core — non-standalone mode — which delivered speed but none of the slicing or low-latency features. Standalone (SA) deployments with a true 5G core are now the majority of new builds and are what unlocks the enterprise capabilities above. The next step, 5G-Advanced (3GPP Release 18 onwards), adds AI-driven radio management, better positioning accuracy, support for "RedCap" devices that sit between smartphones and simple sensors, and improved energy saving. This is the network that will be live through the late 2020s.

Toward 6G

Standardisation of 6G is targeting commercial deployment around 2030. The research themes are consistent across vendors and institutes: sub-terahertz spectrum for extremely high capacity over short ranges; integrated sensing and communication, where the radio network doubles as a radar that maps its environment; AI native to the air interface rather than layered on top; and integration of terrestrial networks with low-Earth-orbit satellite constellations so that coverage becomes genuinely global. Direct-to-device satellite messaging is already shipping on consumer phones; 6G aims to make that seamless rather than a fallback.

The lesson from 5G applies: the technology will arrive roughly on schedule. The business models, the sites and the applications will take longer — and the value will show up first where an enterprise controls the whole stack.