Technology markets have lurched forward in large dollops. From about 1980 through the early 2000s, it was all about the PC. With Apple’s introduction of the iPhone in 2007, focus shifted over the smartphones. Next up: the Internet of Things (IoT), which is really a catch-all for “everything else.”
The IoT market is still quite young, unsettled, fragmented, not at all set in stone. However, some major lines have been drawn — rather quietly — in the past few months, giving us an inkling of how the market will look.
By now, it’s clear that one of the biggest, earliest, most-coherent IoT markets will be automobiles. Cars are already a rich nest of sensors and computers. Over the next few years, these assets will be brought into a connected network that includes the autos themselves, devices people inside them are using, other cars on the road, the road itself, urban infrastructure like traffic lights, the cellular network, satellites, other objects on or near the road, and services of all sorts. The market is coherent because there are relatively few players at each level: a dozen or so automobile manufacturers and about the same number of technology companies.
Other IoT markets include (but are not limited to) smart home, smart cities, retail, manufacturing, oil & gas, non-auto transport, healthcare, consumer products, and advertising. These and others are less formed and will involve potentially many players.
But which companies will win, and who will own this market is taking shape right now.
Intel (and Microsoft) owned the PC market. Intel’s strength in powerful processors that executed complex instructions gave it an edge in a market that remained performance-hungry for several decades. But Intel was in the wrong position for smartphones, the next big market. ARM, with its architecture based on shorter instructions, was better placed to supply a computing platform where low electrical power consumption was more important (at least initially) than raw performance.
Intel never wavered in its view that it had the right architecture for just about anything, that all it had to do was continue the march (which it was leading) to smaller geometries, and that the problem would solve itself. Except that didn’t happen. The smartphone market reached maturity without Intel’s participating at all. There is even some question as to whether Intel can continue to lead in process technology, as it appears to have delayed its first 10nm processor for the third time. We’ll see.
When IoT became a thing in people’s minds, the whole tech industry began to line up for the next big opportunity. Intel launched a series of pilots that could have become product lines, but were withdrawn not long after, science projects hung out to dry. The company discontinued its Joule, Galileo, and Edison product lines, disappointing the Maker crowd, who had pledged their allegiance to the ARM-based Raspberry Pi. In Maker forums I perused, I found a large number of comments that Intel had never really supported the IoT platforms with tools, documentation, and software. They were science projects, but not very will funded ones.
Earlier this summer, Intel shut down its entire wearables division, which had experimented with smartwatches and fitness trackers, with CEO Brian Krzanich saying the company would thenceforth focus on augmented reality. Then, last week, Intel canceled Project Alloy, a merged reality initiative that made use of Intel’s RealSense depth-sensing and Movidius computer-vision technologies.
It might not surprise you to learn that when I contacted Intel’s analyst relations manager in charge of IoT, she told me the company would not have anything to say on the subject. Think of it: a person whose whole job is to represent the company’s IoT effort to the analyst community saying “no comment.” Well, you can see why that might be the case. The landscape is strewn with the wreckage of Intel’s early IoT initiatives. If you don’t have anything good to say, don’t say anything at all.
So, to read the tea leaves without help from the medium, I examined Intel press releases, statements from partners, news stories, and other sources. Conclusion: Intel is not done with IoT by any means, but it is abandoning the endpoint devices, all those little nodes of which there will be billions before everything is said and done.
Aside from Intel’s efforts with Mobileye — an Israeli company that the Santa Clara silicon giant picked up earlier this year to give it a place in visual processing for autonomous vehicles — the company’s main focus in IoT will likely be on what are now called “IoT gateways.” At the moment, this type of product is embodied in Dell’s Edge Gateways for IoT. These units play to Intel’s strengths. High performance architecture that runs hot: they are plugged into a wall, so low power isn’t an issue. Manufacturing scale: they are fairly indifferent to the type of data passing through them, so they can be deployed across many applications. Well entrenched technology: they act as an edge to existing server farms, almost all of which run on Intel products.
But here is where it gets interesting. Although the topology of IoT networks is not entirely baked yet, several aspects of it are clear enough. With all those billions of sensors out at the periphery of the network throwing off huge quantities of data, no one wants that data to make the round trip from field to cloud and back again. So, clearly, there will be some local processing done not far from where the data is gathered.
A very simple example would be a temperature sensor that samples 10 times per second. That’s a lot of numbers to send to the cloud if really all you care about is when it goes over 104° F. Something smart close by could watch those numbers and, when one goes over that limit, send a single message upstream saying, “breach detected.”
The question is, whether that smarts will be right on the proverbial light pole, in a small processor nearby, or in an Intel-style gateway somewhere closer to the cloud. Right now, it looks like Intel has retreated to the gateway, but the ARM camp wants to put intelligence on or near the periphery.
Thus, the battle lines for the IoT are drawn, however fuzzily for the moment. Intel wants the edge gateways, which could be a decent size, reasonably homogeneous market, and the ARM guys will take all the rest, which will involve billions of low-power, low-cost chips in a wide variety of applications. Of course, the ARM folks also has their eye on servers (and therefore gateways), but here the odds-makers favor Intel, which is entrenched. Be that as it may, it’s clear — at least for now — that Intel has given up on the IoT volume market.