Market forecasts tend towards hyperbole and over optimism because that sells reports and so we often find that hyperbolic projections are constantly being deferred year by year until either the sector concerned fizzles out completely or finally really does take off. We have been seeing this with small cells ever since they first arose over a decade ago and are now starting to be deployed at scale after years of such deferred projections of growth. The same holds for the IoT and now the sub-sector of that called the industrial IoT (IIoT).
Such thoughts sprung to mind when the latest IIoT forecast hit our in-tray from Juniper Research, with the headline prediction that the number of IIoT connections would more than double in volume from 17.7bn in 2020 to 36.8bn in 2025. It also insisted that the manufacturing sector would contribute 22bn of these, or about 60% the total, with private 5G networks playing a pivotal driving role in that sector.
We agree with those last two points but might question the headline focus on connections which may be more appropriate when considering other aspects of connectivity such as mobile backhaul. By IIoT connections Juniper is presumably including all links, which at the moment are mostly WiFi and fixed Ethernet in manufacturing, while LP (low power WAN) features prominently in other sectors, especially agriculture and energy/utilities.
It is probably a mistake to merge all these sub-sectors into one overarching use case called IIoT because they have little in common other than scope to benefit from connectivity at various levels of performance, capacity and latency. Some sectors will be largely served by LP WAN and the question then comes down to relative penetrations of the various options, both the two 5G-associated ones of LTE-M and NarrowBand IoT (NB-IoT), along with non-cellular alternatives like LoRa WAN and Sigfox. These vary in bandwidth, range, cost and latency, and we have covered these in some detail in previous bulletins.
LPWAN will also play a role in manufacturing, but in this case as part of larger wireless strategies around 5G. Ethernet delivers the required capacity, bit-rate and low latency, but its wired connections are at best restrictive in the case of factory process control for example and at worst impossible for UAV (Unattended Vehicle Control) and robotics.
Some manufacturers have deployed WiFi to avoid wires, but especially before availability of the latest sixth generation of the technology that has failed to meet targets for low latency, scalability or robustness for those more demanding use cases that only 5G promises to satisfy. Such a swing to 5G though will not necessarily drive up the number of connections that quickly but merely increase capacity and bit-rate while cutting latency compared with WiFi, or for that matter 4G/LTE in the relatively small number of instances where that has been deployed in manufacturing.
On the revenue front Juniper does assert somewhat enigmatically that software will account for 80% of IIoT value in 2025, or $216bn, which sounds rather high. Again though such a blanket prediction overlooks the disparity between sectors and presumably hinges primarily on expectations in smart manufacturing where machine learning is set to play an increasing role over the next few years for various processes, including predictive maintenance and process automation.
At this early stage though many manufacturers are still conducting 5G trials and/or trying to assess whether it will justify the investment of migrating from incumbent Ethernet or WiFi networks. Some of these trials have been quite substantial, involving simulation around digital twin replicas of processes or complete factories, an example being the 5G-SMART project with its trial facility at the Fraunhofer Institute for Production Technology (IPT) in Germany set up by Ericsson.
Such projects will not necessarily convince manufacturers that 5G will automatically prove more cost effective than a well-planned and managed WiFi 6E network incorporating the latest enhancements. Similarly, a greenfield factory site might still decide Ethernet using the latest more compact and lower cost components will serve fixed assembly lines and processes with the required low latency and high performance, without being an encumbrance if the cabling is intelligently integrated into the fabric.
Even for the greenfield, the fact many industrial components do not yet have in-built 5G connectivity could also be persuasive in favor of WiFi or Ethernet. For existing manufacturing sites in automotive and other industries where there is already a lot of connectivity, this could impose considerable friction against 5G migration. Being the best connectivity solution alone will not win many converts without support in sensors or other components, so the industry needs to exert pressure on that front alongside 5G roll out. It may still be that a period of parallel running will be required during what could be quite a lengthy transition phase.
There is then the question of whether to go private, with growing momentum towards that among some sectors, especially automotive. Arguments used in favor include security and control, as well as guaranteed performance. Operators will have to convince enterprises that with the aid of network slicing they can deliver all these benefits more scalably and affordably over their own infrastructures, even if that still involves unlicensed spectrum.