00:00Why is it right for you to go from autonomous driving into the world of robotics?
00:05Well, Mobileye is an AI company in the field of autonomous driving, driving assist, the full spectrum of computer vision and AI to control cars.
00:18And humanoid robotics has been recognized in the past two or three years as a complementary domain.
00:24And Mobileye wants to, it's not only another growth engine, but from a technological point of view, if you are an actor in this area of physical AI, you want to then extend it to the full scope of physical AI.
00:39And I believe that robotics, humanoids, they have a great future.
00:43Now, maybe it's a long-term play.
00:44It's not, you know, a year from now, but I believe it has a significant, exciting future.
00:51And from Mobileye, there's synergetic areas in infrastructure, in AI, in software, in AI talents.
01:00So from Mobileye, this is a great move.
01:02I mean, of course, in many ways, people would say you'd think that because you helped found Menti.
01:08Yeah, that's true.
01:08Your son's work at Menti.
01:11Yeah.
01:11Help me get comfortable with that because many have thrown such accusations Elon Musk's way sometimes in previous acquisitions and not liked it.
01:19Well, it's called a related party transaction.
01:22Has been done in the past and will be done in the future.
01:25And there are ways to handle it, like being recused from decision, which I did and so forth.
01:30And, you know, my son has a 0.00001%.
01:35He's just an employee, so it's not a big deal.
01:37You know, he graduated from computer science and he wanted to do some work and go work for a company that is very, very strong in AI, working on the future of physical AI.
01:50So it makes sense.
01:50But, you know, that is really immaterial.
01:53The material part is the scope of AI.
01:59You know, physical AI includes humanoids, includes autonomous driving.
02:02It makes a lot, a lot of sense.
02:05Now, Mobileye, over the years, was always looking for a new growth engine.
02:08For example, take computer vision and apply it maybe to security areas in cameras and so forth.
02:15But we never found something that really clicked, found something that has significant modes, that has a significant time.
02:21And humanoids is becoming that click, is becoming this area where you can grow.
02:27You and I have been talking for years about, you know, Mobileye's existing footprint and then it's go-to-market in the future.
02:34What's the go-to-market for Mentee?
02:36Well, we see the go-to-market in two phases.
02:39The first phase is structured environments like warehouses, assembly plants, retail, where the number of tasks are finite, are also understood in advance.
02:52What are the tasks that the customer is interested?
02:54A customer will buy a fleet of robots, not just one robot.
02:59And you can customize per customer.
03:02That, I think, based on POCs that Mentee has been doing, if you look at the demonstrations that have been showing, like robots picking boxes, moving them from place to place.
03:12You would essentially just sell the hardware?
03:13Sell a robot, either sell or lease a robot.
03:17We believe that the manufacturing cost in a reasonable volume of tens of thousands of robots would be $20,000.
03:24So this gives a lot of flexibility in a business model.
03:28So this is phase one.
03:29We believe it will start in 2028.
03:32Phase two is unstructured environments like home use.
03:35That is much more complicated because the tasks are open-ended.
03:40Right.
03:40You need to continuously learn.
03:42So it's not that you can prepare yourself to all the possible tasks in a home use.
03:47So you need new technology.
03:50So it's not the same technology you use for structured environments.
03:53And this is one of the things that I showed in my keynote, a technology that Mentee, and this is why Mentee is called Mentee, from Mentory.
04:00The robot is watching, passively watching a human showing a new task.
04:05And the video goes up to the cloud.
04:08In the cloud, there's a foundation model that takes the video, moves it into a simulator, trains over the simulator.
04:14Today, it's a few hours.
04:15In the future, a few minutes.
04:17It goes back to the robot, and the robot performs the task.
04:19Let's go back to your core business, Amnon.
04:21This week, NVIDIA showed us a full-stack solution for autonomous driving.
04:26You know, they make arguments that their software and hardware competencies will make them win.
04:32Your evaluation of what they're offering against what you plan to do, cost for mile basis, how worried you are?
04:38I'm not worried.
04:39Competition, I believe, is always good.
04:42It kind of ignites the market.
04:45In my keynote, I had Christian Zenger, the CEO of Volkswagen Autonomous Cars, chairman of Moya, come to the stage and talk about 100,000 vehicles in the next eight years.
04:57100,000 robotaxis in the next eight years.
05:01Talked about that we are almost there in removing the driver, going driverless in Q3.
05:07Yeah, but are they using Mobileye?
05:08Yeah, of course.
05:09Ironclad?
05:10Ironclad.
05:11There are 100 IDBus vehicles in multiple locales.
05:16All of them is Mobileye stack.
05:19And we are really on the way of removing the driver in a very cost-effective solution.
05:27Our sensor set is, you know, $10,000, $12,000.
05:32Compute is very, very cost-optimized.
05:36Mobileye chips are one-tenth of the cost of competing chips.
05:41We developed our imaging radars, which are the best on the planet, those imaging radars.
05:46So I feel very confident that we are on our way to a new growth engine with the robotaxis.
05:51Well, you're striking deals.
05:53We understand there's a U.S. automaker that you struck a deal with for 9 million cars for safety software.
05:58Who is that?
05:58Okay.
06:00U.S. automaker.
06:01There are three, four.
06:02No, pick one.
06:03Oh, come on.
06:03It's the name.
06:05It's not simple.
06:06But I think that this 9 million units is a big news because we are talking about the evolution of ADAS.
06:14So ADAS today is a front-facing camera.
06:16Maybe you have a front-facing radar.
06:19This 9 million units is called Surround ADAS.
06:21You're taking a front-facing camera, four parking cameras, multiple radars, a very strong value proposition to the customer, and it replaces multiple ECUs in the car.
06:33It replaces the ECU of a parking, ECU of a driving monitoring system, hands-free driving on highways, safety functions for the next five years.
06:44So this is the evolution of ADAS, and it's high volume.
06:47We have two customers in the past year, Volkswagen and this U.S. customer, 19 million units.
06:53This is big.
06:55AI21, we're hearing that NVIDIA, Mr. Huang, is quite interested in another one of your companies.
07:02He's here this week.
07:03What can you tell us about how talks are going?
07:05Did you manage to speak to Jensen while here?
07:07You know, the Israeli press is not as tight as the U.S. press.
07:11So don't believe anything you read in the Israeli press.
07:13So no talks with NVIDIA?
07:14No, there are talks.
07:15I talk to NVIDIA.
07:16I talk to others, but no, it's nothing even close to talking about it in the press.
07:22How do you think Mobileye investors would feel about an acquisition of AI21 by another player?
07:30It has nothing to do.
07:32AI21 is working on foundation models and agents.
07:37It has nothing to do with the activity of Mobileye, activity of Mentee.
07:40Mobileye is physical AI.
Comments