Mobileye has built a multi-billion-dollar business supplying automakers with computer vision technology that powers advanced driver assistance systems. It’s a business that last year generated nearly $1 billion in sales for the company. Today, 54 million vehicles on the road are using Mobileye’s computer vision technology.
In 2018, the company made what many considered a bold and risky move when it expanded its focus beyond being a mere supplier to becoming a robotaxi operator. The upshot: Mobileye wants to compete directly with the likes of Waymo and other big players aiming to deploy commercial robotaxi services.
TechCrunch sat down with Amnon Shashua, Mobileye’s president and CEO and Intel senior vice president, to find out why and how — yep, acquisitions are in the future — the company will hit its mark.
First, a bit of history.
The company, which was acquired by Intel in 2017 for $15.3 billion, took the foundation of its advanced driver assistance technology and pushed toward the development of a self-driving vehicle system. In September 2018, Mobileye announced it would launch a kit that includes visual perception, sensor fusion, its REM mapping system and software algorithms. But developing, testing and ultimately supplying autonomous vehicle technology is a very different business than running a robotaxi service for consumers.
Mobileye has made four agreements to test robotaxi services, including in Jerusalem, Paris and China. Last week during CES 2020, Mobilieye announced it would launch a robotaxi pilot in Daegu City, South Korea. All of these have a human safety driver behind the wheel.
The move may seem like a distraction for a company, but Shashua insists it is the right strategy.
The following interview has been edited for length and clarity.
TechCrunch: Why are you doing this? You already have a robust business.
Amnon Shashua: If you look at the business layers of Mobileye, we have ADAS, which is very well-defined. We are a Tier 2 supplier; we sell a chip — the chip comes with all the software, hardware and distribution.
If we continue vertically, the next level is to build a self-driving system — a complete end-to-end hardware plus software plus validation kit. You’re talking about 360-degree sensing of cameras, you need to add radars and lidars. It’s not only sensing, it’s the driving policy of decision making, it’s the control of the vehicle. It’s the validation guaranteeing something about the probability of an accident or probability of a failure because there’s no driver in the loop. We’re talking about safety-critical issues which are exponentially more difficult than driving assist.
There was realization that dawned on us awhile ago. The Holy Grail of this business is passenger car autonomy: where you buy a passenger car and you pay an option price and with a press of button it can take you autonomously to wherever you want to go.
The realization is that you can’t reach that Holy Grail if you don’t go through the robotaxi business.
Why is that?
The cost of a self-driving system, for one. The 2022 range is somewhere between $10,000 to $30,000 per vehicle [in addition to the base cost of the vehicle].
We’ll be close to $15,000 when we launch in Tel Aviv to 2022. That’s just the hardware sensors, the PCB (printed circuit board), the chip.
That actually seems lower than I would have expected.
Yeah, our competitors are somewhere around $40,000 or $50,000 per vehicle.
When you think about it for a robotaxi, a $15,000 investment is OK. Even $30,000 is okay. You can still make a very good business with this kind of capex investment. But for a passenger car, this is too much. If this is the cost, the carmaker needs to make multiple on it, so then it will come to many many tens of thousands of dollars per car. Even for a premium car, that’s outside of the scope.
Then comes regulation. It’s much easier to regulate an operator that operates a fleet of cars than a consumer. Operators have reporting responsibilities, back-office responsibilities, teleoperation responsibilities. There’s someone to contact if something goes wrong. With the consumer, it’s much more complicated.
The third issue is scalability. If you sell a car for a consumer, the promise is that you can go everywhere. It has to go everywhere, otherwise the consumer will ask, “why did I pay so much money?”
To support this, there needs to be high-definition maps that are everywhere and refreshed continuously because they have to be accurate. This is something that doesn’t exist today and it will take a number of years. We have this technology and we’re launching it — this is our REM technology with crowdsourced data — but it takes time until it’s ubiquitous.
Then comes the issue of homologation. If I’m promising you that you can drive from everywhere to everywhere. Say you are in New York and you want to travel to San Francisco — and with the press of a button it will take you there. You’ll need to pass through many, many cities and each city has its own culture of driving, so I need to be able to collect data and to experiment in many, many places. I can’t experiment in San Francisco and then launch it in Boston.
This becomes hugely expensive.
Making this investment without having a business model to generate money while I’m going through this huge investment would kill any company.
If I need to make this investment and if I need to put vehicles almost in every city, why not generate revenue from that? This is where the robotaxi comes in. It’s a necessary phase no matter how you look at it.
You could go through partnerships.
Yes, but when you sell a self-driving system to someone then there’s the issue of who gets the data. And then there is the the huge investment of the homologation. If you’re not generating revenue from the service, it could be too much.
Then comes the realization that when you look at ride hailing today, reaching profitability is very challenging. Look at Uber, look at Lyft.
If you do the calculation right, the driver is about 80% to 85% of the economical value of ride hailing. Our calculations show that the cost per mile for ride hailing is 30% lower than a taxi. If you remove the driver with our system, it will be 50% lower than that. If now you do ridesharing, so you have more than one passenger in a car, it’s about 30% to 50% lower than that.
The impact is really game-changing because the cost of ride hailing today is equivalent to the cost per mile of a light user in a metropolitan area like the center of San Francisco. If I’m a heavy user I’d better own a car. If I’m a light user, it’s better to take an Uber or Lyft and not to own a car.
But now if you look at the cost of a robotaxi with ridesharing for a heavy user in a metropolitan area and in a suburban area, it’s worthwhile to use the service and not to own a car. Not only that, it rivals the cost of a bus.
What you’re describing reminds me a bit of how you run the rest of your business. And I get you want to control the data, but managing a ride-hailing fleet is a totally different business than what you’ve done before. It can be very costly and complex. Dealing with human beings can be costly, and getting people to share, which is important for unit economics and avoiding congestion — is the crux.
Did you hesitate at all and wonder what the fleet management side was going to be like? Will it require Mobileye to build out a whole new segment within the company? Can it?
So we’ve been working on it in house for the past year and a half. There’s a fleet management system, there is route optimization, a mobility intelligence that that you need to have, there’s teleoperation and the back office. … there are so many layers that you need to have and it’s very, very complicated.
We are working with partnerships with many small companies, each one specializing in a certain area of the layers of the service. We’re also growing teams of people who are developing other areas that we have decided we’ll do ourselves and not partner with others. And we will consider favorably inorganic growth as well.
Inorganic growth?
Yes, this is acquisitions. We are into Intel, it’s not just Mobileye, we are Intel. We can acquire in a big way.
So we were looking at it, you know, very seriously both from building partnerships and acquisition of all the layers of service. We calculated the size of the opportunity to say that within a decade the market size of a robotaxi would be close to $160 billion. So it is a serious business.
And if we look at other tech companies, if you look, like, at the Apple, you know, many tech companies are including to their product, also service. It is something that you see more and more tech companies also doing.
And we are doing partnerships. In Tel Aviv, the partnership with the Volkswagen and Champion Motors, it’s a joint venture. In Paris we’re doing this with our RATP. With China, we’re doing it with the Nio, which will take our self-driving system and selling cars where our software system is built in.
The Nio deal seems more like that Holy Grail moment that you are kind of describing.
Yes, but it’s in China. China’s easier.
The first, I’d say 10 cities (globally), there are going to be many partnerships.
How long will it take you to get to 10 cities?
We start in 2022 and we have already four, so I’d say by 2023 there will be 10 and then we’ll ramp up from that much faster.
How can any company (like Mobileye) compete with the network that Uber or Lyft has created?
Well, we will be starting small initially and it will be geofence so we won’t be competing directly at first. And then, as I said, we’re looking favorably into inorganic growth as well.
How much attention are you going to place on acquisitions? And in what areas? Ride-hailing or fleet management? What are the hot spots?
This, I cannot get into, but I hope that within not a long period we could make some interesting announcements in those areas.
