Intel recently announced it is ceasing its efforts in the small chip (Joule, Quark, Edison) and maker board market (Curie), which were programs aimed at establishing Intel’s presence in the emerging Internet of Things (IoT) marketplace. With all of the activity currently taking place in the IoT world, does this signal that Intel has given up on this space? Not really.
These programs were run out of Intel’s new technology group and not out of its recently restructured and reemphasized IoT group. This change is less a surrender than it is a strategic regrouping for Intel. In fact, I see this as a wise move as its unlikely Intel could compete at the low end of the “things” market with the ARM camp, and especially with the likes of Qualcomm, MediaTek and others that can leverage their strengths and scale in mobile phone SoC designs into IoT. The mobile phone processor market is an area Intel was not successful in impacting in any significant way (other than in supplying modems), and with its very limited scale, it would be hard for Intel to profitably compete in the processor/SoC space with these high volume providers who are already producing tens of millions of chips and can relatively easily re-purpose a portion of these to the IoT segment.
That’s not to say Intel is totally abandoning the IoT market. It will continue its efforts within its IoT group to focus on compelling higher performance compute platforms that can leverage its investments in Atom, Core and Xeon chipsets, as well as its capabilities in modems. Intel plans to target markets like autonomous vehicles/automotive, drones, healthcare and similar areas where higher end and more compute intensive options are needed and where lowest cost is not the primary driver. Although lower volume than “gadgets,” this industrial and enterprise space has much better margins and characteristics that fit into Intel’s plans and strengths. And it’s an area where Intel has significant intellectual property to exploit.
Data is the issue in making IoT succeed
With a typical autonomous car generating 4TB per day and a plane generating 40TB, it’s clear that data analysis to optimally run these systems will not be done on a small circuit board independently within the individual sensors and/or actuators. It all needs to be connected in a feedback mechanism that can generate actionable intelligence, often through AI and/or high-level data analytics. IoT doesn’t work without the back end processing services necessary to run the systems taking the actions. Smart cities, autonomous driving, healthcare, industrial systems, energy/utilities are all “sensoring up.” Yet unless the data can be collected and properly analyzed to make informed decisions about what needs to happen next and enable the proper actions, what’s the point? And while “edge” computing that puts processing power at the point closest to the sensor is required as a data pre-processing point at the “things,” only 1 percent or less of the raw data will actually be transmitted into the cloud. It’s clear that some “beefy” compute capability is required at the back end.
But what is Intel’s strategy for IoT? In a few words, It’s the back end that counts!
Most of the money to be made will be in back-end services built around analytics, AI/ML, big data processing, network operations, etc. This is not just about CPU and/or GPUs, but also about storage and connectivity. There will also be a need for edge computing that includes some significant computing requirements (e.g., under-the-hood autonomous driving computing platforms and large scale infotainment for cars, visual/LIDAR signal processing systems for autonomous driving, and also in areas like healthcare, and real time analytics of large data sets sent from many sensing points).
This is where Intel maintains an advantage in that it can leverage many of its current enterprise/business assets to deliver services itself and especially through cloud providers (e.g., Google, Microsoft, AWS) that have deployed Intel platforms. Large data center processing elements that include optimizations for enhanced data analytics, AI and ML will be critical and Intel plans to lead in this space through its current investments (e.g. Xeon, Nervana, Movidius, FPGAs), and future acquisitions (e.g., MobilEye). The edge compute platforms will be more competitive as they may be built into the SoCs that run the “things,” but I see a multilayered approach to edge services that include significant stand-alone requirements.
Bottom line
Even a company the size of Intel has limited resources, and being in too many areas of potentially low growth (by Intel’s measurement) markets makes little sense when those resources can be targeted at more lucrative and strategic markets. That’s what Intel is doing, and it’s a good move on their part. Intel is content to let others fight for and potentially dominate at the front end processing nodes, and this is a positive market development for Qualcomm and MediaTek among others. Intel has its eye squarely on being a leader in the back end compute services needed to make the full promise of IoT a reality, and its confident it can succeed in this space in a big way. I wouldn’t bet against them.