With autonomous cars all the rage at CES 2017, it wasn't hard to find test vehicles using Mobileye's (MBLY) technology.
BMW, which announced at CES it plans to deploy 40 autonomous test cars in 2017 that feature Mobileye and Intel (INTC) processors -- it's aiming for a 2021 commercial launch -- supplied one. As did auto parts giant Delphi (DLPH) , which Mobileye is teaming with to create a self-driving platform to sell to automakers in 2019.
I talked with Mobileye Chief Communications Officer Dan Galves about how his company, the biggest supplier of vision processors for driver-assistance (ADAS) systems, is positioned in the budding autonomous driving hardware market. And what the company considers its main strengths relative to the competition.
Galves pointed out Mobileye's EyeQ 5 processor -- it's expected to begin sampling in the first half of 2018 -- is powerful enough that only one processor will be needed to power autonomous driving systems.
But in the interim, solutions will rely on multiple Mobileye EyeQ 4 processors; the EyeQ 4 is currently sampling, and expected to be production-ready by year's end.
Importantly, Galves estimates Mobileye's average selling price (ASP) for an autonomous driving solution relying on all of the company's relevant technologies -- i.e., processing for eight cameras, surround view, driving policy algorithms and mapping -- would be in the $400 to $500 range. That's roughly 10 times above the $45 ASP Mobileye currently records for individual ADAS solutions such as lane-departure warning and collision avoidance. A Level 3 system, which provides partial autonomy, is estimated to yield an ASP about 3 times that of a single ADAS solution.
On how the autonomous driving solutions Mobileye is working on compare with those of Alphabet's (GOOGL) Waymo self-driving unit, Galves emphasized Mobileye is going with a "sensing-heavy, map-light" approach to autonomous driving. A car's cameras, radar and lidar sensors create a model of a car's environment, and maps are mostly there for redundancy.
Waymo, by contrast, appears to depend heavily on HD mapping. Galves argues such as an approach is capital-intensive -- Google, of course, has a long history of investing heavily on maps -- and also raises challenges with regards to keeping maps updated and accurate. "It doesn't take much for the vehicle to get to a point where its lidars are not matching up with what the maps say they should see," he said.
Galves also stressed there has been a sea change in how automakers look at autonomous driving: They now see it as an opportunity to transform their business models into ones featuring recurring revenue streams, where (via autonomous ride-sharing fleets) access to cars is sold by the mile. As a result, automakers are intent on not getting left behind, and eager to partner with whichever technology suppliers can help them bring autonomous cars to market sooner.
On the subject of Mobileye's own self-driving partnerships, Galves says his company is open to teaming with numerous firms, and that existing alliances, such as the ones it has with Delphi and mapping provider Here, shouldn't be considered exclusive. "Delphi has some unique qualities, because they have very strong software for planning, behavior, driving control... Our technologies fit really well together, but we're open to working with other partners as well," he said.
When asked what Mobileye viewed as its major competitive advantages relative to autonomous driving processing solutions from rivals such as Nvidia (NVDA) and NXP Semiconductor (NXPI) , Galves quipped "That we have software."
Though admitting other chip makers can produce competitive silicon, he adds Mobileye sees itself as a software company at heart. "That's really where most of our value is... From that perspective, we really don't feel like we compete with the hardware suppliers," he said.
Nvidia announced autonomous driving alliances with Audi and Mercedes-Benz at CES, as well as with major auto parts suppliers Bosch and ZF Group. Mobileye has ADAS design wins with Audi and Mercedes.
On the subject of Tesla Motors (TSLA) , which is using Nvidia's Drive PX 2 module in its Autopilot 2.0 system after relying on Mobileye for its predecessor, Galves seemed to choose his words carefully. He pointed out there are similarities between what Tesla is shipping and the systems Mobileye is working on -- both use eight cameras and radar -- but he added that Mobileye's solutions, by having multiple radar devices along with lidar, will have more redundancy.
He also observed Tesla appears to be developing proprietary autonomous driving software that runs on top of Nvidia's hardware -- the company just hired Apple software exec Chris Lattner to head its Autopilot software work -- and that Mobileye can rely on all of the video data taken in from test work done by clients to improve its own software. Galves: "Tesla is a very unique, very capable company. Our customers don't seem to be interested in trying to develop their own proprietary software when they can use Mobileye's."
Mobileye's autonomous driving push has come at a price: The company's R&D spending rose 48% annually over the first 9 months of 2016 to $47.2 million, nearly matching revenue growth of 50%. Galves notes Mobileye wants to bring its R&D spend down to a level of about 9% of revenue over the "medium-term," which compares with a recent level of about 15%.
However, he adds Mobileye isn't going to keep its R&D spending below what it considers necessary for the sake of hitting a target. "We'll spend what it takes to execute the products that we need to execute."