Dan Levy: And then wanted to just follow-up on the conversation specifically on Chauffeur. So it sounds like AMAS there is maybe more of a focus on dedicated platforms, less retrofitting, more partnership, etc. Chauffeur, maybe you can give us some sense. I know that was part of the engagement conversation that you mentioned, but you know, how significant is your spend on Chauffeur right now? What is the interest in Chauffeur are your customers seeing this as sort of an evolution of SuperVision so it is very aligned with the SuperVision spend, or is this a separate stream and it is something that maybe the timing is getting pushed out a bit more and that is playing into the OpEx commentary?
Amnon Shashua: It is very aligned with the SuperVision. I think of it as kind of an incremental addition to SuperVision. SuperVision is mostly camera based. There are some radar it as an option. For example, in the ZEEKR 001 there is a front facing radar. In the Porsche program, there is also surround the radars. And when you go to eyes off the Chauffeur, you are adding some ladder as well in order to create more redundancy and create and a bit more compute. Instead of two EyeQ 6 that we have in the Porsche program, it is a three EyeQ 6. So it is really, it is really incremental. The heavy lifting is not so much on the development, it is on the validation, because you need to prove that you are multifold times better than human statistics, crash statistics and that creates an effort of validation this is something that we are working together with the OEMs. We are creating hardware in the loop farms of thousands of ECUs for each program.
For example, for the S662 [ph], for the CH663, for the DR64, each one has a hardware in the loop farm of many, many thousands of ECUs in order to run two thousands of hours of data for per night. And this is ongoing and part of our budget, part of our OpEx growth. It is not something that we did not anticipate or would come as a surprise. In terms of the OEM traction, we are in serious engagement with a number of OEMs, I believe that at least two of them will be able to close this loop.
Dan Levy: Great. Thank you.
Operator: Our next question comes from Adam Jonas with Morgan Stanley. Please go ahead.
Adam Jonas: Hi. Thanks everybody. Amnon, what are your thoughts on the advantages or disadvantages of using custom silicon versus GPUs such as an NVIDIA A100 for SuperVision training. Curious what mobilized strategy is regarding custom versus GPU and is there any effort to move towards a custom system in a vertically integrated way the way some of your competitors are?
Amnon Shashua: Our system is vertically integrated. We have an EyeQ chip, but instead of GPUs, we have our own accelerator families. We have five different families of the accelerators, and that would make our chip very efficient. If you look at the SuperVision two EyeQ chips now on 001. On paper, the total tops is 30, something compared to that one tenth of the tops on paper of the competing solution, and you don’t see any advantage in terms of performance for the competing system. So we have highly efficient solutions. And the advantage of a highly efficient solution is cost, power consumption, size of the ECU, whether you need to – have you need to call it. Power is very important when you are talking about a vehicle. So our approach, which is not a general purpose chip like the A100, it is really customized to the type of, you know, workloads that we need in order to power both computer vision and driving – has great advantages of efficiency.
Adam Jonas: Thanks. I will leave it there. I will hold a follow-up. Thanks.
Dan Galves: Thank you so much. And thanks everyone for joining the earnings call. We will talk to you next quarter. Thank you.
Operator: This concludes today’s conference call. You may disconnect your lines at the time. Thank you for your participation, and have a great day.
