Get email delivery of the Cadence blog featured here
Mike Demler gave the keynote at the Linley IoT conference a couple of weeks ago. He is a senior analyst there, and also a senior editor of Microprocessor Report and Mobile Chip Report. His background is in analog and mixed-signal design at companies like TI and GE. And Cadence. In fact, he worked for me in marketing years ago back when I ran the (then) Custom IC Division.
At the Embedded Vision Summit earlier this month, Chris Rowen used the two graphs below to point out how fast "deep learning" has gone from not even on the Gartner Hype Cycle chart to being at peak hype. Meanwhile, in 2014, IoT was at peak hype, and in 2015, it is still there. Every time I blink, the number of "things" seems to have gone up a billion or two. Mike started by putting it into perspective: 25B things means 3.5 per person. Some of these are PCs, servers, smart TVs, smartphones, and so on, not new "things" but established markets. The focus of the IoT conference is new things, incremental applications and emerging markets, including embedded systems that were previously not connected. So iconic examples are smart electrical power meters and smart thermostats like Nest.
Most IoT devices connect to the internet through gateways. They may be wired but most are expected to be wireless. IoT devices sense or control (or both) some part of their environment. There may be specialized IoT hubs in the middle, that Cisco calls the "fog". Beyond that is the cloud, and that is standard datacenter hardware with no differences due to IoT.
Three big market segments for IoT are industrial, consumer and wearables.
Industrial IoT includes smart buildings, cities, hospitals, farms, and retail. The value proposition is often improves process efficiency, reducing costs and, perhaps, doing things that were not possible before. Right now, this is the largest segment but a lot of that is due to smart meters that make up nearly two-thirds of the segment with 300 million out of 2B already installed. Smart buildings have potential for another 2B sensors, smart farms have major savings in water and labor (and, sometimes, three rice harvests per year versus two, for example). Smart factories offer opportunities for increased automation but also for scheduling maintenance that used to happen only when something broke.
The consumer IoT market is growing slowly. In the long term, it has the potential to be larger than industrial, but there is not killer app right now. Consumers only achieve marginal savings from things like smart lighting or thermostats, but the cost is so much more than dumb versions that only the affluent buy. Purchases are motivated more by convenience and status. Devices only used for content consumption, such as smart TVs, are excluded.
Wearables is growing faster but is low margin. For example, the Xiomi Mi band, comparable to a $100 Fitbit, is just $15. Apple dominates in smartwatches with about 12 million of the 24 million shipped, but it is still a fraction of the smartphone market, which is measured in billions. Fitness band are nearly two-thirds of the market, with smartwatches making up most of the rest.
I have no idea if the technology is possible, but there is promising work on on-wrist blood pressure monitoring based on different colored LEDs. Something like that might be the killer app, or rather the keep-alive app. Who wouldn't pay a few hundred dollars for half-a-day's warning of a stroke or heart attack?
Mike talked about wireless interfaces, what he called "the IoT air wars". I thought I knew a lot about wireless interfaces but he had lots I'd never heard of. The summary would be that there are lots of wireless interfaces, and no one interface is appropriate for all situations. Having said that, there are clearly too many standards and I doubt that they will all survive in the long term.
The other technology trend is towards standard products. Vendors such as Atmel, NXP, Silicon Labs, STM, TI, and Toshiba are adding radio. Vendors strong in wireless such as Broadcom, Marvell, Mediatek, and Realtek are adding programmable processors. Basically, IoT requires a combination of a low-power processor and connectivity, so if you have one then you need to add the other.
The other big thing you need is security. It wasn't the theme of Mike's presentation but it was a theme that other presenters came back to time and again during the day. A low-power processor and connectivity is not enough, you need security. More and more that means a combination of hardware and software, since software-only solutions have proved to be inadequate.
So what's the forecast? The graph above is Linley's current outlook. Slow maturation in consumer and wearables pushes the peak growth out to 2018/19. It takes until 2018 for IoT consumer to overtake industrial (in units), and until 2022 to overtake smartphones (again, in units).
William Golding, the screewriter, is famous for saying that in Hollywood, "nobody knows anything" meaning that nobody can tell in advance which movies will be hits. I get the impression that IoT is like that. You can take the current market segments (watches, fitness bands, smart meters, fire alarms...) and run the numbers but a new Walkman or iPod may come along and the world will look different.
The IoT is growing rapidly, but initial estimates were unrealistic
The short conclusion, the agenda for the rest of the conference: key features are connectivity, low power, and security. Especially security. Did I mention security?
The next Linley conference is the mobile one, that I see is now Mobile and Wearables, to be held July 26 and 27. Get the details.
Next: What Is RocketSim? Why Did Cadence Acquire Rocketick?
Previous: It's HOT in Austin In June