TOKYO — Mobileye is aggressively shedding its reputation as a one-trick vision chip supplier. As the technology of automotive industries transitions from driving assist to robo-taxi development, Mobileye, an Intel company, is rolling out a comprehensive plan to grab a sizable share in the highly automated vehicle (HAV) market.
Mobileye’s newly disclosed roadmap ranges from a silicon-only “open” EyeQ5 chip (which allows third-party codes to run) to a complete subsystem focused on perception, a turnkey robo-taxi hardware system and applications for ride-hailing businesses.
Mobileye’s strategy to move up the value chain from chips to systems is reminiscent of Intel’s CPU strategy in its early days. Intel succeeded in the PC business not just by selling its own CPUs but by designing and marketing PC subsystems and eventually PC motherboards.
In an exclusive interview with EE Times, Amnon Shashua, Mobileye’s CEO, explained, “Mobileye is unique in a sense that it is the only company” offering an “open” strategy, allowing “room for collaboration” with tier ones and OEMs.
Shashua said that Mobileye competitors such as Nvidia and NXP provide silicon with which carmakers can write their software. Meanwhile, Waymo, Uber and GM Cruise are building closed systems that they don’t sell to anyone else.
Mobileye’s difference, he concluded is that “we build our own car, end to end, while offering options for complete systems, subsystems, chip and software, and silicon alone.”
Since last month, when Bloomberg reported that Mobileye is opening its driverless technology, EE Times began investigating what exactly Mobileye is “opening up,” and what it means to the industry.
First, a little background.
While Mobileye’s EyeQ series chips have a stellar reputation in the Advanced Driver-Assistance Systems (ADAS) market, the company’s strategy is not exactly popular. Mobileye’s proprietary vision solutions are based on tightly coupled EyeQ chips with Mobileye’s own perception software.
Mobileye’s rivals describe the EyeQ chip series as a “black box.” Mobileye wannabes have told us that carmakers are scrambling to find a way out of Mobileye’s stronghold in perception algorithms. They want EyeQ alternatives onto which they can plug their own software and add their own “secret sauce.”
As OEMs crave freedom from Mobileye’s proprietary approach, the question is how long Mobileye can afford to stay with its black-box business model. As the era of autonomous vehicles — Level 4 and Level 5 cars — approaches, what must Mobileye do, and what exactly is the company’s plan for the upcoming EyeQ5?
This why EE Times wanted to talk to Shashua.
Click here for larger image (Source: Intel/Mobileye)
‘Closed’ and ‘open’ EyeQ5 chips
Shashua said during the interview, “We haven’t really changed our colors” since its first EyeQ5 announcement .
EyeQ5 is Mobileye’s newest SoC fabricated by using TSMC’s 7nm process technology. It is expected out next month, according to Shashua. At a launch announcement in early 2016, Mobileye described the new SoC as “offering the vision central computer performing sensor fusion for fully autonomous driving (Level 5) vehicles.”
Shashua confirmed in our recent interview that besides the “closed version of EyeQ5” with Mobileye’s proprietary silicon and software tightly integrated (and nobody can change it), Mobileye is now rolling out a “silicon-only version of EyeQ5,” for the first time.
Shashua explained that this is akin to Intel’s Xeon chip for laptops, PCs and servers, on which “other people write software.” He said, “This is what Nvidia does today.”
Mobileye’s silicon-only approach to EyeQ5 marks a stark contrast to an EyeQ business model in which it sells “silicon and software as a closed system.” Shashua said that in driving assist the closed EyeQ chip comes with “entire application detecting pedestrians, vehicles and whatever it needs to function in a closed system.”
So how does “open EyeQ5” work for real customers?