This story appears in the June 29, 2016 issue of Forbes. Subscribe
Mobileye cofounder Amnon Shashua is betting that crowd-sourced, high-definition mapping technology will speed the introduction of autonomous vehicles. (Photo Credit: Ronen Goldman for Forbes)
It takes Amnon Shashua 20 minutes in city traffic to get to his Jerusalem office each morning–not a terrible commute, especially since his AudiA7 handles most of the chore. “I let go of the wheel and let the car drive,” says Shashua, 56. “It’s really fun.” Fun, but not entirely worry-free. “I have to be alert,” he adds.
As cofounder, along with Ziv Aviram, 57, of Mobileye, a leading provider of assisted-driving software, Shashua knows better than anyone that autonomous cars still need work. At highway speeds, under certain conditions, hands-free driving is already doable, as Tesla MotorsTSLA -1.35% has shown with its Autopilot system, which, like others from Audi and Cadillac, relies on Mobileye’s camera-based software. Where it gets dicey is in complex situations that arise on city streets, like merging into a roundabout.
Artificial intelligence, which is moving faster than anyone in the auto industry expected, is helping with those situations. But it’s still likely to be another decade before cars are smart enough to drive themselves at all times without any human input.
Mobileye, two years after its $890 million IPO, has a clever plan to speed things along–one that’s quite different from that of Alphabet, the parent company of GoogleGOOGL +0.14%and Mobileye’s biggest rival in the race toward a self-driving future.
Mobileye EyeQ software chips are used by more than 90% of the world’s automakers, helping cars stay in their lanes and brake in emergencies. Now, working with three of the biggest carmakers, the company is rolling out new high-definition mapping technology that will use those chips to gather crowdsourced, real-time data from vehicles and pinpoint a car’s location in relation to traffic signs, lane markers and other objects. Together with cameras and other sensors like radar, the continually updated maps will provide an additional source of information that should help Mobileye spread automated driving features to more cars more quickly.
“Mapping is a way to lower the bar” for driverless cars until artificial intelligence is ready, says Shashua. General MotorsGM -0.86% will get the technology first, through its OnStar connection, followed by Volkswagen and Renault-Nissan in 2018. Together the three automakers sell 30% of the world’s cars, providing the large numbers Mobileye needs to make crowdsourced maps work. Other global automakers are lining up to add the technology, Shashua says.
Mobileye’s route is different from and will likely be quicker than the one Google is taking with its moon-shot strategy. Google wants to leap directly into the self-driving era and has gone so far as to develop prototypes that have no steering wheel or pedals. These and other specially rigged cars have logged more than 1.5 million self-driving miles to perfect the technology.
But the tests are limited to roads Google has meticulously mapped and annotated near its Mountain View, Calif. headquarters and in three other cities. These maps are far more detailed than the GPS navigation maps in today’s cars and smartphones. While GPS maps are accurate to within 10 meters or so, high-definition maps measure details within a few centimeters, recording everything from the height of a curb to the width of an intersection to the exact location of a stop sign, creating a 3-D portrait of the car’s surroundings. Google cars navigate by comparing data collected from various sensors, including a spinning laser scanner mounted on the roof, with those recorded 3-D maps. These laser-based radar systems, called lidar, are bulky and expensive but are getting smaller and cheaper.
The bigger problem is that those detailed scans of streets chew up a lot of bandwidth. An hour’s worth of 3-D data collection is equivalent to 20 years of individual smartphone usage, according to Rod Lache, an analyst at Deutsche Bank. While Google’s mapping capability is “an amazing thing,” Lache says, it’s hard to scale. “There are 4 million miles of roadways in the United States, and one-third of them are not even paved,” he says. “How will Google–or anyone that wants to do this–scan all of them?” Even then, there would be the additional challenge of keeping the maps up to date.
Google doesn’t seem too worried. “We’ve mapped the world–twice,” a spokeswoman says, referring to Google Maps and its Street View capabilities. All the data Google’s test cars need to navigate are stored in an onboard computer, but that could change someday. Google’s focus for the moment is making sure the technology works, she says.
Mobileye’s breakthrough was finding a way to create 3-D maps using the camera-based chips already installed in millions of cars. Its Road Experience Management software can identify landmarks and other selected roadway information at extremely low bandwidths (approximately 10KB per kilometer of driving) because it doesn’t record every grain of detail, as Google does. Though not quite as precise as Google’s 3-D maps, Mobileye’s software, when combined with camera images and other sensors, can determine a vehicle’s location to within 10 centimeters, making automated driving possible.
Mobileye, founded in 1999, has two trends working in its favor. Virtually every car will have a front-facing camera within a few years because the auto industry, working with U.S. safety officials, has agreed to make automatic emergency-braking systems standard on all vehicles, and those commonly need cameras.
In addition, automakers want to provide over-the-air software updates for their cars, just like makers of computers and smartphones do. By 2022, 98% of new cars in the U.S. will be connected to the cloud, according to IHS Automotive, opening up the data pipe Mobileye needs to crowdsource its maps. “The beauty,” Shashua says, “is there are other incentives for having communications and cameras in the car.”
With an approximately 90% market share in assisted-driving systems and innovations like 3-D mapping, it would seem that Mobileye is well positioned for the driverless future. The company’s IPO in July 2014 was the biggest ever for an Israeli company. In 2015 it had $241 million in revenue, up 68%, and net income was $68.5 million. Free cash flow nearly doubled to $96 million.
Yet its stock now trades near $40 a share after peaking at more than $64 last summer. Investors worry about increased competition from new rivals such as semiconductor maker Nvidia, which recently unveiled chips with artificial-intelligence features.
Lache thinks such skepticism is unfounded. “In the tech world people are used to competitors catching up very quickly,” he says. “In autos it’s a lot more challenging. I don’t think people understand the complexity of what Mobileye has accomplished.”
Ironically, the looming threat from Google might be just the thing that secures Mobileye’s success as a partner to established automakers. Many are afraid Google wants to control people’s digital lives, whether at home, on their phones or in their cars. Auto companies aren’t willing to yield that relationship with their customers.
“We’re at a point where automakers aren’t looking for the lowest price. They’re looking for the leader,” Lache says. “Looks to me like Mobileye is part of their defense strategy.” The best offense, after all, is a good defense.