Intel's (INTC) 2016 product announcements, M&A activity and layoffs have all had a grand, overarching theme to them: The chip giant no longer expects the PC CPU market to be a growth driver, and is instead focusing most of its resources on supplying processors and complementary chips for data center hardware and embedded devices.
It makes sense that this effort includes a push to grow Intel's presence within cars -- the largest embedded chip market if one excludes mobile phones, and one whose size is steadily growing thanks to several trends. But it's also a hotly competitive market in which automakers hold a lot of sway, and as Intel's latest automotive announcement shows, one in which the company will sometimes need a helping hand from fellow chipmakers.
Five months after Intel announced a deal with BMW and top automotive vision processor supplier Mobileye (MBLY) to help BMW bring a self-driving car to market by 2021, the company has announced it's teaming with Mobileye and auto parts giant Delphi (DLPH) to provide self-driving systems for automakers in general. Mobileye and Delphi had already inked a deal in August to develop a self-driving system that they promised would be "production ready for 2019."
In addition, Intel announced on Tuesday that it's creating an Automated Driving Group (ADG) to spearhead the development of self-driving solutions; the unit will be headed by former IoT Group chief Doug Davis. The move comes two weeks after Intel declared its VC arm (Intel Capital) will make over $250 million in autonomous driving-related investments.
Interestingly, Intel says the solution it's working on with Delphi and Mobileye will initially feature a CPU from Intel's high-end Core i7 line, and will eventually use a more powerful processor set to be unveiled in a few weeks. As of November 9, the cheapest Core i7 CPU had a list price of $303.
If Intel was able to ship a Core i7 or something comparable within 5 million cars annually (a small fraction of the estimated 75 million cars due to ship this year), that could spell a $1.5 billion-plus opportunity, not counting any sales of complementary parts such as flash memory or communications chips.
At the same time, the fact Intel has needed to team with Mobileye for both the BMW and Delphi deals suggests the company still lacks the vision-processing expertise to single-handedly provide the brains for self-driving systems. Mobileye, whose EyeQ processors have gone into tens of millions of driver-assistance systems, has a lot of experience developing chips and algorithms that make sense of what a car's cameras are seeing in real-time. And these chips have taken in a lot of real-world driving data that should prove valuable as Mobileye develops algorithms for fully-fledged autonomous driving.
From the looks of things, the plan is to pair these abilities with the raw computing power of Intel's CPUs, with the end result being a system that can immediately process driving data from numerous sensors (camera-based or otherwise) and handle any other function needed for a car to drive itself. It won't be cheap at first: Delphi exec Glen De Vos says the system will initially cost about $5,000 per car, but adds prices will likely fall quickly as adoption grows.
There could also be an opportunity for Intel on the back-end: The company's Xeon server CPUs could ultimately be used to analyze driving data from millions of self-driving systems, and thus help make the algorithms that power them more intelligent. Xeon Phi accelerator cards featuring technology developed by AI chip developer Nervana Systems, which Intel recently acquired, could also play a role here.
But in both fields, GPU giant Nvidia (NVDA) looms large. Nvidia's Drive PX 2 module for autonomous driving systems is being tested out by numerous automakers, and is used by Tesla Motors' (TSLA) second-generation Autopilot system (Mobileye is the supplier for the original Autopilot system). The Drive PX 2 combines two standalone GPUs with two of Nvidia's Tegra app processors; Tegra chips are already widely used in automotive infotainment systems.
Nvidia's GPUs have been quite effective for many AI/deep learning algorithms, and the company is banking on this ability to help it win over automakers. It also makes Nvidia Intel's biggest rival for server-side processing of autonomous driving data; the company's Tesla accelerator cards have been widely adopted by cloud service providers for AI-related workloads.
NXP Semiconductor (NXPI) , set to be acquired by Qualcomm (QCOM) , is also a notable rival; the company is the auto industry's largest chip supplier, as well as Mobileye's biggest rival. Earlier this year, NXP unveiled Bluebox, an autonomous driving platform that pairs a vision processor with an eight-core "embedded compute processor." In time, NXP could offer solutions that leverage Qualcomm's Snapdragon processors and connectivity chips.
Between stiff competition and the fact the autonomous driving market could still be several years away from taking off, it's best not to get too excited over Intel's deals with BMW and Delphi. Like a slew of other recent moves, the deals and announcements are signs Intel's management has its priorities right, but a number of hurdles still need to be crossed before there's a significant payoff.