Mobility As a Digital Lifestyle Product | Nakul Duggal (Qualcomm) & Pranay Agrawal (Fractal)

I think back then for us it was actually pretty straightforward. We wanted to diversify into the automotive industry. We felt that there are not many industries where you're building 80, 90 million units every year. An industry that is not as exposed to everything that is going on in tech. It was just odd from my point of view. Why would you have an industry that is not at par with where the rest of the world is? And so the approach that we took was we're going to be a technology partner to the automotive ecosystem. And there were clearly many leading thinkers in the automotive space that were happy to partner with us. General Motors back then, Audi. And the approach we took was we have to spend time first earning the right to be heard. So we trusted advisor. It wasn't so much about building a business. It was about can we build trust with this ecosystem and can we shape the road map? Can we shape our products and our technologies to be relevant to the auto industry? Because the pace of change is different. The quality requirements are different. It takes a long time for the automotive ecosystem to trust a new supply chain. The risks are much different. So ultimately it came down to us looking at the market and saying we can make an impact here. There is value for us here and for them. And the ecosystem looked at us and said, 'Are these guys going to be around for a long time or is this more opportunistic for them?' And I think when you start to spend time with a platform where you start to think about the potential of the platform, what became pretty clear was that the problem statement was the same whether you went to Korea or Japan or the US or Europe. The problem statement was exactly the same. Everybody wanted to get more out of the platform but the industry was just not exposed to what was possible. So the approach that we took was to create our own definition of what will be possible. We invested in partnerships with Google for example in the early days we said hey why don't we create an Android version of what would it take for this experience to be relevant in the car. We took a lot of our chips and we said we must make these chips very specific to what is needed for a cockpit versus for a smartphone. We spend time on what does it mean to build products that meet the quality and reliability levels that a car is going to need because this product is going to be on the road for 15 years. You can't just treat it like a consumer device. And when you do all of those things at the engineering level, that requires investment. It requires a concerted push and you have to do it at scale. As we started to take that approach, a lot of other things started to fall into place, right? Smartphones became very relevant. People wanted to have this consumer experience. What was very important for the auto industry was to not get relegated to where when a consumer sits inside their car, they immediately allow the cockpit of the car to be taken over by the smartphone. Automakers hate the idea that a smartphone will just kind of take over this very important asset. So, it was not difficult to convince automakers that this real estate that they had is something that they should differentiate on. But they need help. They need to create experiences, curate them. And that journey kind of frankly it kickstarted something that brought a lot of digital ecosystems closer to the automotive industry and kind of took off. So we've run a lot of business in the cockpit space mostly by enabling automakers and their value chains to be able to experience what is possible when you want to bring these new experiences. And so that I think has gone off quite well. The other thing that we did as we got deeper into automotive was we started to see what was going on in driver assistance systems and driver.

Look, when you spend time within an industry very deeply for many years you essentially want to be able to make sure that you are relevant to that ecosystem. But for us as a technology company more importantly why would we not be able to pivot towards trends that really are here to stay? Reducing collisions by applying technology into a car seems like a fairly straightforward idea why would you not implement it? And what we ran into was number one this is a safety application so this is not a consumer application so when you buy cars for ourselves or for our families, you have to be able to meet a certain bar of quality, a certain bar of compliance. That was something that we felt was important to do because it would give us scale in a market that is not easy for anybody to just enter into. The second piece really was the convergence of machine learning and computer vision and later all of the AI that you now see with hardware. You have to co-design the hardware and AI together for this to work because this is not a cloud application. This is a real time safety critical low latency application. And so we decided okay we have to build chips for this market and ultimately we have to build a stack for this market. And so the journey that we kind of started to go down was we've done enough of connectivity we know how to do the cockpit. Why don't we get into the driver assistance business? And that required us to then double down on building a chip road map which I think the automakers all really liked because they wanted to have somebody who would bring the prowess, the investment, the scale and the stack business which followed. We made an acquisition of a software team that existed within a tier one because we wanted to get a jump start in terms of what does it take to build a trusted piece of software that is responsible for a safety application. I think it was a very good decision because what it did for the auto industry was they looked at us and said here's a partner now a company we didn't know really very well but they become a partner they are investing in this space they are at scale and we need more companies like Qualcomm to help us become a fully digital ecosystem because we are not going to be able to have the scale of the consumer ecosystem and at the same time there's tremendous potential in the car industry. So I think that bet that we made six, seven years ago was a really good one because we supply to every single automaker on the planet today.

You know I think it's an interesting question because when you're in the middle of it you don't really operate like that right you operate with a tremendous amount of instinct with the trust and commitment that you are getting from an ecosystem. What was very clear was that we heard the same thing everywhere. So there was not a single part of the ecosystem that said I don't believe in this. So it really kind of came down to does the ecosystem know what to do. Everybody knew that they were pointed in the right direction. See the auto industry is a business that has a very different way of working than most other ecosystems because the risks are very high if you make a mistake. And so a lot of it really just came down to is there enough trust that we have built with this partner with their technology with their ability to stay they're a consumer company what do they know about automotive is it better for us to just stay on the safe side and go with somebody we've worked with for years but then on the other hand they saw how much was going on in other ecosystems and how can you fall behind. I think the other piece certainly was Tesla right in the US at that time as Tesla was becoming really popular as an EV platform with a very different kind of experience. There was certainly an aspiration around okay we need to be able to build products that are equally exciting. It was a completely different class of product. And then I think the China phenomena happened probably 3-4 years ago maybe mostly after COVID but the rate at which China expanded first it was shock nobody could believe what was going on in China and then obviously global automakers have a footprint in China so they started to internalize and see how different this was. So I feel like for us this was always a long I've been doing this since 2011. So for me one year here or there was not really the point. It was always going to be a long gestation business. We had the ability to really double down and make sure that we were committed to what we were I mean there was never any doubt that we were on the right track. It was just about when would it take off.

I think this whole concept that physical AI where you can embody AI into a platform that is responsible for transporting you from A to B and you can do it in a way where it will do it safely. It will do it better than a human driver. I mean we are on that path right that's the path that is in front of us and I think there will be a time in the not so distant future where you will have robots that will drive cars and cars will not get into accidents. I think to make this happen universally obviously there's a lot of complexity around it there's a lot of heterogeneity around it but it is pretty clear the tech is here now it's a question of when.

Yeah. Well, I think you have to test the heck out of your system is one thing that you have to do. So I think the approach that we take is first of all I think the industry's done a pretty good job in building a lot of regulation around what is the compliance that you must follow when you're developing a stack to make sure that you are not inadvertently going to create a situation that causes an accident. We have two layers of the stack software that we write. The underlying layer is what we call the safety guardrail which are a set of rules that are hardcoded that the vehicle must follow where it cannot whether it's for an automatic emergency braking staying in your lane not hitting an object pedestrian anything that might come in your field of view. And those are very heavily hardened and they form the ground plane. They form the foundation of the stack that we build and we are doing this across a 100 countries. So this is hardcore. It requires a tremendous amount of driving a lot of data collection making sure that you are actually compliant with the letter of the law as might be relevant in that market. Then on top of that you add a lot of AI based stack development and AI based stacks improve and learn as you feed them more data and there what you're trying to do is to make the stack there is now this term that is used in automotive called end-to-end networks where you feed a sensor on one side and out comes a decision on how to navigate what to do as far as the behavior as far as the policy for the trajectory that you're going to take. And as AI becomes smarter, as AI becomes much more capable, what is going to happen is you feed data into AI models and these models will essentially be able to predict what is the right thing for the vehicle to do. So a lot of what you see with FSD is a combination of this naturalistic driving but combined with a lot of nudging of the algorithms in terms of what to do in the real world. Not an easy problem to solve because you need a lot of data, you need a lot of experience and so that's what everybody's marching towards. My expectation is in the next 5 to 10 years you will actually see a lot of platforms on the road where you will have FSD type experiences but frankly it's a matter of time once you have mapped out enough use cases enough corner cases I think you will get to that experience.

Exactly right.