What will replace 2G and 3G for IoT-powered AI?

A range of artificial intelligence (AI) applications depend on connected Internet of Things (IoT) devices. For example, AI automated data analysis can be applied to large quantities of data generated by IoT devices such as environmental sensors and industrial machine monitoring.

IoT devices ‘in the field’ need reliable cellular network connectivity. All eyes may be on 5G, and the possibilities it opens up for faster, more rewarding online experiences and virtual reality (VR), but for many, today’s reality is mass deployed IoT devices that still rely on 2G and 3G.

However, those networks are being phased out, to make way for 5G. That puts companies and device manufacturers at a crossroads, with decisions to be made. They must plan to transition their devices from 2G or 3G to another networking technology.

What is 2G/3G sunsetting and when does it happen?

The closure of 2G and 3G networks is often referred to as ‘sunsetting.’

Unfortunately, there is no single, straightforward answer to the question of when it is happening. 2G will be switched off in many countries by the end of the decade with 3G going even sooner in many cases. The UK government has a target to phase out all 2G and 3G by 2033 and BT has announced that customers across its brands will be phased off 3G by 2023.

Dates and plans vary from country to country and operator to operator. As many IoT-powered AI applications have a global reach, that makes for a potentially complicated picture.  

What companies need to consider

Migrating from one networking technology to another takes considered planning. Companies may need to decommission and change out subscriber identity module (SIM) cards contained in IoT devices to keep them working. They may need to upgrade, or even change out, those devices.

In place of 2G/3G, companies will need a networking technology that is most suited to their devices and applications.

As each new technology standard for cellular networks is announced, we become accustomed to thinking in terms of bandwidth. 5G promises high capacity, high speed connections to download high-definition films in mere seconds and enjoy no-lag VR experiences. Yet, when it comes to the IoT, often only low processing power is required for small data packet exchange.

IoT Managed Service Providers have evolved their network management capabilities.  As TinyML and other AI/ML techniques propagate, there are increased options for network design in deploying solutions. What those IoT devices do often need is coverage. Hence, attention turns to low power wide area technologies for the IoT as we bid goodbye to 2G and 3G.

How to weigh up the cellular alternatives

Low power wide area networks (LPWANs) are a group of wireless technologies for IoT connectivity over wide geographic areas. They include Long-Term Evolution Cat-M1 (LTE-M) and NarrowBand IoT (NB-IoT).

This makes LPWANs potential contenders for companies managing fleets, building management systems or environmental monitoring and many more similar IoT implementations where coverage is a key consideration. Devices may have a low data throughput and use low power, but they will need to remain deployed for many years.

However, when weighing up the pros and cons of LPWANs, there is a snag. Some national markets support one or the other very well but only a few support both. What’s more, it wouldn’t be possible to have a single SIM, used interchangeably for NB-IoT and LTE-M, in multiple countries. That’s a significant drawback because most companies, for logistical simplicity and cost management, will aim for a single stock keeping unit (SKU) with their device.

Depending on the design of your AI-driven IoT network LPWAN may be entirely unsuitable.  For example, latency and throughput are poor in LPWAN.  If you have deployed sensor concentrators, it’s likely that your concentrators are bursty in their output and may occasionally need large data up/downloads for updating and patching.

Added to this, mobile network operators may choose to phase out support for one or other of the technologies if adoption rates don’t justify sustaining it. That would leave companies in the unenviable position of having to transition again; something they will be keen to avoid.

For devices that remain deployed for potentially years, across many market regions, network availability and consistency of operation are significant requirements. For these, and other reasons, LTE CAT-1 (specifically LTE CAT-1 BIS) is an attractive alternative.

CAT-1 is an established 4G service, therefore it’s within every network’s technology stack already around the world.

That makes it a definite contender for large-scale IoT deployments. However, while CAT-1 offers relatively low power connectivity, depending on the application it is likely to offer a shorter battery life than NB-IoT and LTE-M. For context, the latter may last up to ten years while CAT-1 may be around half that. Still a considerable length of time, and not an issue for devices with rechargeable batteries or a connected power source.

There are also cost considerations to weigh up – CAT-1 is likely to be a bit more upfront but roughly the same per megabyte in operation.

Companies making the switch from 2G/3G must weigh up their requirements against the characteristics of the cellular alternatives. Cost, coverage, time to market, power and manufacturing process considerations will all be in the mix.

A range of networking technologies offer a migration path for IoT installations that currently connect through 2G or 3G. There are advantages to each, and all priority requirements must be factored in when deciding on a solution. With the right support from dedicated service providers, device manufacturers and companies with IoT deployments can evaluate the options open to them and make an informed choice to sustain their applications for the future.