As the auto industry wonders what Silicon Valley chip supplier might dominate the market for self-driving cars — giant Intel or colossal Nvidia — the answer may well be headquartered 7,400 miles east of Detroit.
When Intel announced plans in March to acquire Mobileye of Israel for $14.7 billion, the Jerusalem-based software supplier commanded 70 percent of the global market for obstacle detection software.
Obstacle detection will be a crucial new vehicle component in the coming decade, empowering cars to “see” road obstacles such as pedestrians, stray animals, dangerous debris, vehicles and signs. In addition to detecting such objects, the systems must recognize and understand what those objects are doing. Is that a pedestrian in the middle of the lane ahead, or it is an emergency worker warning the vehicle to stop?
This has been futuristic technology until now. But it will soon be a standard component for an industry transforming to autonomous driving.
Mobileye gained dominance with EyeQ3, a software-and-chip combo that relies on cameras that are inexpensive enough for use in small cars. And that may be Intel’s ace in the hole.
To be sure, both Intel and Nvidia are showcasing their top-of-the-line technology in vehicles developed by Audi, BMW, Volvo, Mercedes and other luxury brands.
But Mobileye gives Intel a chance to introduce self-driving technologies — such as lane-keeping, collision avoidance or 360-degree surround view — step-by-step in the mass market.
The company also can generate crowd-sourced road maps from camera data uploaded by Mobileye-equipped vehicles to the cloud.
General Motors, Nissan and Volkswagen all plan to introduce Mobileye maps over the next two or three years.
“Our ability to generate this map in an efficient and low-cost way is a big value to our customers,” said Mobileye spokesman Dan Galves. “Cameras are an extremely versatile data collection device.”
Rock the boat
Once Intel’s acquisition is completed — most likely late this year — Intel will integrate its own chip development with that of Mobileye.
Intel is developing its own supercomputer, dubbed Xeon, which it is expected to market to luxury brands such as BMW. Meanwhile, Intel likely will offer its less expensive Atom processor to automakers for anti-collision features.
At the same time, Mobileye will roll out two new generations of obstacle detection systems — dubbed EyeQ4 and EyeQ5 — capable of analyzing data from a mix of cameras, radar and lidar. Those systems will become available in 2018 and 2020, respectively.
Shashua: Will oversee activities for both Mobileye and Intel.
Shortly after the acquisition was announced, Intel said Mobileye co-founder Amnon Shashua would oversee the activities of both companies. Mobileye’s Jerusalem operations will become the center of Intel’s work in the field.
“Dr. Shashua has a ton of experience in computer vision, and he understands where the industry is going,” Doug Davis, senior vice president of Intel’s Automated Driving Group, tells Automotive News. “Our intent is to take advantage of his experience.”
Thus, Mobileye will form the core of an alliance that Intel is building with other suppliers to help BMW commercialize self-driving vehicles by 2021.
In May, Delphi Automotive announced that it would be a Tier 1 integrator for the project, and Continental AG joined the partnership one month later.
Meanwhile, Nvidia has formed alliances with Autoliv, Bosch, Hella and ZF TRW.
Angelos Lakrintis, a London-based analyst with the consulting firm Strategy Analytics, says these competing announcements by Intel and Nvidia are “the latest stage in the emergence of two autonomous vehicle tribes.”
The two alliances are locked in a seesaw battle for dominance. Nvidia has announced deals to supply its Drive PX computer to Tesla, Mercedes, Audi, Volvo and Toyota. Mobileye will continue to supply its EyeQ obstacle detection package — which includes software plus a computer chip designed by Mobileye and produced by STMicroelectronics — to a variety of automakers.
For the next three or four years, Intel will have an opportunity to pitch those customers on the benefits of its “decision-making” chips — for example, Atom or Xeon — to guide self-driving vehicles.
But Mobileye’s momentum “will slow down with the upcoming rise of faster processors” that are optimized for deep learning, Lakrintis says.
Lakrintis says two serious challengers will be Qualcomm Inc. — the largest global producer of automotive computer chips — and Renesas Electronics Corp., which once held that title.
Qualcomm, based in San Diego, shook up the industry in 2016 when it announced the acquisition of NXP Semiconductors NV for $47 billion.
Last year, NXP introduced its BlueBox module, which supports a variety of collision avoidance features.
Meanwhile, Tokyo-based Renesas has introduced Renesas Autonomy, a computer platform for self-driving vehicles. The first product released for that platform was an obstacle recognition system for cameras and lidar.
Since Renesas and Qualcomm have deep pockets, both can afford to make investments for the long haul. And the prize may be worth it.
Within five years, automakers are expected to reach purchasing levels of $20 billion to $30 billion a year for technology and parts for self-driving vehicles, IHS Markit forecasts. The suburban Detroit-based consulting firm projects global sales of 600,000 autonomous vehicles by 2025 and 21 million by 2035.
Mobileye and Intel will get their share. In March, IHS analyst Akhilesh Kona said the two companies will be “a one-stop shop” for the software and processors that automated vehicles will require.
Kona expects their combined annual revenue from driver assistance systems will total $10 billion by 2022, up from $2 billion this year.
Mobileye’s EyeQ4 obstacle detection system will debut next year in 2 million vehicles.
Component: Obstacle detection software, processors
Function: Identify vehicles, pedestrians, animals and other objects in the road
Description: The system consists of software, chips, cameras, lidar and radar.
Vision: Sensors and obstacle detection systems will be the “eyes” of self-driving vehicles.
Challenge: Mobileye needs crowd-sourced roadmaps using camera images uploaded from millions of vehicles.
Other r&d participants: Intel
Likely market appearance: Mobileye currently produces EyeQ3 for cameras. Higher-grade EyeQ4 and EyeQ5 will appear in 3 years.