AI rollout will determine the outcome.
- Much like versions of Windows, the mobile network generation seems to oscillate between transformation and ho-hum for each generation, with the good news being that on this basis, 6G should transform the digital world.
- Last year at MWC, 6G was far enough away that no one really had to worry about it, but this year, we are within 3 years of launch, and so a great deal more attention is being paid to what it is and what it will deliver.
- Technically, it appears that the 6G standard will be fairly similar to 5G, but with modifications to optimise it to so that it can meet the 3-7x growth in traffic expected by 2034 and deal with how traffic patterns will alter.
- I think that the argument over hardware upgrade cycles is largely over, as radios have been moving to being software-defined for some time now, meaning that 6G should be a software upgrade and not much more.
- This means that the growth of the mobile industry and the equipment vendors that supply it will be a function of traffic growth and patterns, and much less about rip and replace to access a new technology.
- Hence, the question becomes how the traffic in the network is going to change over the next 10 years and what changes need to be made to the infrastructure and devices that will be supporting that traffic.
- Here, the No. 1 topic on everyone’s lips remains AI, and given that there is a strong economic incentive to push inference to the edge of the network, connectivity is going be a crucial element in making sure that it works properly.
- Here, the industry sees four factors that will impact how the network will have to be built going forward:
- First, Network performance: using more AI in the network promises to improve its radio performance.
- This is particularly in areas where the signal is weak, or the environment is complicated, such as in the middle of a city with tall buildings.
- By including this as part of the radio system, higher frequencies become usable, which will be important when looking at ways to economically increase the capacity of the network.
- Second, Connectivity: If AI is going to play a larger role in the digital lives of consumers, then always-on agents will begin to become popular.
- This means that agents on devices will be running in the background all the time and sending data to other parts of the network.
- Historically, the network has been very download-centric (eg streaming video) and billions of always-on agents will change that pattern.
- Enterprises running digital twins of their assets will also start to shift this balance, which will also increase the sensitivity of the service to latency, which will need to be optimised for.
- This could probably be done with 5G, given how similar the two appear to be, but if 6G is merely a software upgrade rather than a rip and replace, then a new standard makes a lot more sense rather than retrofitting the old one.
- Second, Compute: which refers to the fact that AI is very compute-heavy and depending on the nature of the service, it will be needed in different parts of the network.
- Some services will run better on the device, others on the basestation, others in the local data centre and others on the very large datacentres currently under construction.
- Furthermore, depending on the device being used and the network conditions at the time, the location of the compute will vary.
- The network will need to be aware of this and be able to manage traffic and decide where the most effective place is to run the compute.
- Third, Sensing: this is a relatively new idea that is being proposed by a number of different players in the mobile industry.
- The idea here is that the network’s condition and how the signals are being transmitted and received contain information about the environment, which is relevant to the service being provided.
- For example, drone detection and guidance or the management of robots in a factory are all use cases that could make use of the network being used as a sensor.
- I don’t think that anyone would use this in isolation, but it offers another independent sensor which, combined with lidar, radar, cameras and so on, can be used to reduce the error rate and make the system work better.
- The net result is that I don’t think that 6G is going to drive a typical investment cycle, but will be rolled out as and where it is needed to support the roll-out of AI services across the network.
- This means that its success is tied to AI, meaning that if there is a large correction and services take longer to be rolled out, it could be delayed somewhat.
- However, at the moment, with AI compute capacity being built at a breakneck pace, the outlook for 6G to hit its 2029 target looks to be on track.
- Hence, the way to view 6G is not as hype or just another standard but as an evolution of the network to allow it to work with the new services that AI promised to bring.
- Time will tell if the even number really is lucky.
