Recently the Linley Group held their IoT Hardware Conference. Mike Demler kicked off with an overview of the IoT market, and on the second day Linley Gwenapp presented a deeper dive into IoT chips (leaving out processors available as IP blocks). Cadence announced their new processor and presented on the whole Tensilica audio family (see Meet the Newest Member of the Tensilica Family, the HiFi 3z DSP ). The IoT Market The IoT market is not really a market. But the big division is between IIoT, the industrial internet of things, and consumer IoT. In industrial are all the "things" that support smart buildings, smart cities, smart energy, smart factories, shipping, medicine, and more. Consumer is smart homes, connected appliances, automotive, and wearables. Mike presented some market size data, which was driven by the numbers of homes, factories, and so on. First, smart homes. There are about 800M residential broadband connections worldwide, growing to 1B by 2022. Even in the US, 39% of rural homes lack fixed broadband access. The primary networking environment for smart home devices is expected to be a Wi-Fi router, which is why it is good to start from homes that have broadband access. Depending on your definition, IoT can include smart devices for content consumption (that's a long-winded way of saying TVs). They are explicitly excluded as IoT devices, as are smartphones and PCs. These are devices that ship in big volumes, so big that they are already treated as separate markets in their own right, not as part of IoT. Next, smart (non-home) buildings. There are 30M commercial buildings worldwide. Linley estimates 2.5B client devices, mostly low-cost MCU/sensor/radio with some higher end devices for surveillance with video. These will be used for fire alarms, access control, maintenance, parking management, and HVAC/lighting. Automotive has 1B cars in use worldwide, with about 90M more being sold each year. However, only 12.5M of those were connected cars in 2016 (meaning things like OnStar). Mike pointed out that FCC has allocated the 5.9GHz band for intelligent transportation systems (ITS), which includes vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communications. However, these V2x technologies will not be that significant in the autonomous driving universe, at least in the short term. Too few vehicles will have the technology in the next decade, and I doubt that most traffic lights will be converted to communicate with vehicles in much less than a decade either. So my belief is that autonomous vehicles really will be on their own for the foreseeable future, meaning that vision (and lidar and radar) will be the key technologies (humans drive mostly with their eyes, too). Now here's a number you probably don't know and probably never have thought about before. There are about 3B cattle, goats, pigs, and sheep worldwide. There are big advantages of tracking them, although I think "smart sheep" is an oxymoron (and, by the way, did you know that oxymoron is an oxymoron, with oxy meaning smart and moros meaning moronic?). Perhaps more significantly, there are about 50M corporate farms worldwide (so not counting ranchettes and subsistence farming, neither of which are likely to be installing much IoT infrastructure). The TAM for the farms (on top of the animals) is about 1B devices. There are 10M factories worldwide. There is lots of opportunity for control/monitoring of manufacturing processes, and tracking work-in-progress and finished goods. Our own industry, semiconductor, is completely automated. There are increasing amounts of "additive manufacturing", which is what is more casually called 3D printing. The TAM is about 150 million smart sensors. FItness bands are the most popular wearable devices, selling 3:1 against more expensive smart watches. There are 2M smartphone users but Linley expects only 10% penetration of wearable, so 200M. There are other smaller volume markets for specialized applications, like Google Glass for enterprise. The above graph summarizes the TAM for IoT client devices. Note that these are measured in units and not dollars. As for the installed base, Linley estimates it as 1.6B connected IoT devices, with nearly 60% being in industrial (IIot) with smart meters being the biggest segment (since these are largely installed at private homes, I'm not sure if this should count as home or industrial—although generally the consumers don't buy meters themselves, of course). Technology Trends Almost every IoT device has wireless connectivity of some sort. As a result Low-Power Wide Area (LPWA) networks are key. Although some devices will connect through wireless routers, the current standards are too high power and the data rates are overkill (the new HaLow standard addresses that, but currently installed Wi-Fi routers don't support it). There are other standards too...in fact, way too many. NB-IoT, 5G, LoRa, Sigfox, LTE-M, some of which are hub-based and some of which connect to cellular networks (using specialized protocols, not the regular smartphone protocols). Bluetooth has also been extended to longer ranges and to add mesh capability (where packets can hop from device to device to without requiring a hub, such as lightbulbs and lightswitches). One issue with the LTE standards for IoT devices is that they will have some sort of subscription fee associated with them. Even if it is just $1/month, say, it builds up over a couple of years to a significant sum if the devices only cost $25-50 in the first place. The table below gives the current deployment status of many of these technologies: IoT Forecast Previously the Linley Group forecasts assumed big run-rates by 2019. The numbers are still 1B in 2019 but the peak has moved out to 2021, rising to 2.3B in 2023. Limited infrastructure, lack of standards, and cost have restrained deployment. The predictions are an installed base of IoT devices rising to 10.4B in 2023. Penetration goes to 13 devices per middle-class household. As costs drop, consumer will dominate in unit counts, with 72% of shipments by 2023, industrial 28% of the market. Segmenting the IoT Processor Market In the afternoon of the second day, Linley Gwennap gave his view of the IoT processor market. However, he excluded embedded processors and only talked about actual chips that integrated processors and communication. In effect, it was the trailer for the full-length movie A Guide to Processors for IoT and Wearables . Since most readers of this blog are more likely to be interested in embedded processors than chips, I'll give you the high-level stuff and let you buy the report if you need all the details. Linley's conclusions about these integrated chips that are available: Integrated Wi-Fi processors connect directly to home router, some are multi-mode with Bluetooth, too IEEE 802.15.4 MCUs minimize power and cost but typically require a hub (MCU vendors lead this market) LTE works with things that move and doesn't require configuration, but does have monthly fees (smartphone processor vendors lead this market) New product and vendors emerge all the time Eclipse...Not Just an IDE Anymore I hope you are somewhere to enjoy the solar eclipse today. I'll be at Hot Chips (watch for blog posts in a couple of weeks) and they even have a special break scheduled into the agenda. Sign up for Sunday Brunch, the weekly Breakfast Bytes email.![]()
![]()
↧