Life after 2G and 3G: choosing a network for a device that has to last ten years
A device designed this year may still be on a wall in 2036, and the network is the part of it you control least. How LTE-M, NB-IoT, LTE Cat 1 bis, LoRaWAN, satellite and Wi-Fi compare, and the questions that decide between them.

- Insights
- 25 October 2026
- 5 min read
For twenty years the answer to "how does it connect?" was 2G, because 2G was everywhere and the modules were cheap. That answer has expired. South Africa's 2G and 3G networks must be off by the end of 2027 (Connecting Africa). In the UK, operators began switching off 3G in 2023 and have told the government they will not run 2G or 3G past 2033 at the latest (Ofcom). A device you design now will outlive both, and any module that falls back to 2G is falling back to something that is going away.
There is no single replacement. There are six serious options, and each one is the right answer for some products and an expensive mistake for others.
The options at a glance
| Network | Data | Moving devices | Power use | SIM | Typical fit |
|---|---|---|---|---|---|
| LTE-M | Around 1 Mbit/s | Yes | Low, with power saving | Yes | Trackers, alarms, devices that need regular updates |
| NB-IoT | Tens of kbit/s | No | Lowest of the cellular options | Yes | Meters and sensors in basements and boxes |
| LTE Cat 1 bis | Up to about 10 Mbit/s | Yes | Moderate | Yes | Mains-powered devices, multi-country products |
| LoRaWAN | A few kbit/s, small messages | Limited | Very low | No | Farms, estates, mines and campuses |
| Satellite | Small messages | Yes | Moderate, costly per message | Service plan | Sites with no tower in reach |
| Wi-Fi | Plenty | No | High | No | Buildings whose network you control |
What each one is good at
LTE-M is a slimmed-down version of 4G built for devices. It hands over between cells, so it works on things that move, and its power-saving modes let a battery device sleep for most of the day. It is the closest thing to a general-purpose successor to 2G, and it carries a firmware update of a sensible size without fuss.
NB-IoT is narrower and slower, but better at reaching into basements, meter boxes and underground chambers. It was not designed for devices that move, so it suits things that stay put and send small amounts of data. Firmware updates over NB-IoT are slow, so design the update process before you choose it.
LTE Cat 1 bis is ordinary 4G at a modest speed with a single antenna. It works on any 4G network without the operator enabling a special IoT feature, and it roams like a phone. It uses more power than LTE-M or NB-IoT, so it suits devices on mains power or a large battery. Where LTE-M and NB-IoT support is patchy, or the device crosses borders, it is often the safest choice.
LoRaWAN is a long-range, low-power radio in unlicensed spectrum. Devices send a few dozen bytes at a time, batteries last for years, and there is no SIM. You run your own gateways or use a network provider's. One gateway on a high point can cover a farm, an estate or a mine. It is a poor fit for large messages, for anything that needs an answer in under a second, and for frequent firmware updates.
Satellite now reaches places no tower does: reserves, pipelines, boreholes, the sea. Low-earth-orbit services and satellite versions of NB-IoT carry small messages, each at a higher price than on a terrestrial network. Some modules can now use the cellular network when it is there and the satellite when it is not, which makes satellite a sensible backup rather than a default.
Wi-Fi costs nothing per device and carries plenty of data, but only where you control the network. Where you do not, it is a support problem waiting to happen: passwords change, routers are replaced and guest networks block devices.
5G RedCap, a lighter version of 5G for cameras and wearables, is arriving too. For most fleet devices it is still early.
The questions that decide it
Does it move? If yes, LTE-M or Cat 1 bis. NB-IoT and LoRaWAN are for things that stay put.
What powers it? Mains power opens every option. A battery that must last years points to NB-IoT, LTE-M with power saving, or LoRaWAN, and to firmware that keeps the radio off most of the time.
How much data, and how often? A reading once an hour fits anything. Images, audio or large firmware updates rule out NB-IoT, LoRaWAN and satellite as the main link.
Where exactly will it sit? Underground and inside metal boxes favour NB-IoT. Open country with no towers points to LoRaWAN with your own gateway, or to satellite. The only reliable way to know is to test at the real sites, on more than one network, with the antenna and enclosure the product will ship with.
How many countries? A device sold in South Africa and the UK needs the right bands, approvals and roaming agreements for both. Some networks do not allow roaming on LTE-M or NB-IoT, which pushes multi-country products towards Cat 1 bis, or towards an eSIM that can switch to a local network profile.
How long must it last? Ten years can span two changes of network technology. Choose a modem family with a long supply commitment, keep the radio on a part of the design you can replace, and build in remote firmware updates so the device can follow the network.
Where we start
There is no universal answer, but we do have a starting position. For a device on mains power, or one that moves, start with LTE-M and use Cat 1 bis where LTE-M coverage is thin. For a battery device that stays put and sends little, start with NB-IoT and test the worst location first. For a dense site you control, use LoRaWAN with a cellular link out. Add satellite only where there is no tower. Whatever you choose, prove it in the field on a handful of units before you commit to a thousand.
If you are choosing a network for a new device, or replacing one that is about to lose its network, send us a brief, or write to hello@handee.co.za.
Choosing a network
Pick a network for ten years
Pick a network for ten yearsTell us where the device will live and how long it must last. We will narrow the options before anyone commits to a radio.
- 01A working slice of device, cloud and dashboard, validated against real hardware
- 02Senior engineers who own the architecture and every decision
- 03An honest view on what to prove first, and what can wait
What happens next
- 01We read your briefWe come back to you with the questions that will shape the prototype.
- 02A first callWe talk through the problem, the hardware, the data and what success looks like.
- 03A clear proposalScope and cost for the prototype phase, so you know exactly what you're committing to.