Qualcomm’s Visionary Leap in Automotive

Olivier Blanchard:
Welcome to the session of The Six Five Summit. The title today is Qualcomm’s Visionary Leap in Automotive: A Comprehensive Overview with Nakul Duggal. I’m Olivier Blanchard, Research Director of the Futurum Group, and I’m joined by Nakul Duggal, who is Group General Manager of Automotive, Cloud Compute, and I believe Industrial IoT, is that right?

Nakul Duggal:
That’s right, Olivier. Good to be with you.

Olivier Blanchard:
All right. Good to have you. So I have some few questions for you. We only have 15 minutes, so I’m going to try to get through them a little bit quickly. So obviously, we’re talking about automotive… And congratulations, by the way, on the massive design pipeline that you guys have earned. I’ve been watching your progress very, very carefully for the last few years, and it’s kind of amazing what you guys have pulled off, so congratulations. So, like I just said, $45 billion design win pipeline, which kind of strikes the point that innovative leadership in times of technological disruption also applies to automotive. And so since I have you here, I want to ask you first of all, whether Qualcomm and your strategies for leading effectively in an era of rapid technological change, but focusing specifically on fostering innovation and adapting to do new business models in the automotive sector. Because it’s not just about getting better cars out to market, it’s much broader than that now. So if you could provide some color and some strategy, that would be great.

Nakul Duggal:
Yeah, I think it’s a decade long strategy. This isn’t something that we stumbled upon, and it’s a complicated market. It makes decisions at a well-defined cadence, and yet at the same time, there is just a tremendous amount of change, a tremendous amount of competition that is taking place in automotive. So maybe to start off with from the top, the key ingredients obviously are to have a technology roadmap that is critical to being able to differentiate, and we are fortunate that we span so many different technologies given the portfolio of products and markets that Qualcomm is in, from every type of wireless to every type of compute to every type of AI. A ton of software, a ton of enablement, a very wide ecosystem of partners that we work with. And we do this at an annual cadence. So one thing that is very important in a market like this is that you have to be agile, you have to be able to operate at a global scale, you have to be able to touch so many different industries and technologies.

And the car is becoming this platform that is going through a massive consumer look and feel. Consumers are deciding the types of features they want. These features are coming from technologies that come from various types of industries. So when you find yourself in the middle of all of that, you have to be able to balance between how to move quickly while at the same time being able to deliver products that are designed for this market. So I think we did a number of things. I think we invested in safety early on. We realized that the car is not a phone, the car is a car. The car requires its own unique capabilities. It needs to leverage from ecosystems that are fast moving, but it also has to be able to deal with the complexity of what the car ecosystem is. And we have been doing this now for quite some time. So what we benefit from is we have a massive footprint in the US, in China, in Japan, and Korea, now in India, Europe, obviously, we’ve been working with for a long time.

These partnerships are very deep. They go back a couple of decades, in some cases, we operate across the wireless ecosystem. We operate across the cockpit ecosystem massively. We have been investing heavily in the driver assistance space from a silicon perspective. We are now developing our own stack. We partner with a variety of stack providers. So when you’re in the middle of so much change, when you have a massive portfolio of assets and you know that there is a tremendous amount of competition, being relevant is very important, and being able to watch what is around the corner, what partnerships are emerging, what the competitive challenges are in this industry with China, and the EV onslaught with the US and European automakers, making sure that they can stay ahead. We find ourselves in this very unique vantage point, and it’s about seeing a lot of different things and then acting pretty quickly. I think that’s what allows us to stay ahead.

Olivier Blanchard:
Yeah, yeah, definitely it feels like you’re getting at a cart but also like the full stack of solutions, just building it pretty steadily and pretty quickly at a pretty decent clip. So that’s been really impressive. All right, onto a different topic, slightly artificial intelligence for automotive. I want to know how you think of artificial intelligence and how it’ll revolutionize the automotive industry. Specifically, I think how it’ll enhance vehicle experiences, not just safety but also the driver and user experiences and also build new opportunities for revenue for what driving means for some value that can be created by the automotive industry. And also, I guess, see if you have any predictions or insights on what industry leaders and consumers should anticipate. I have some ideas on that, but I want to hear yours.

Nakul Duggal:
Yeah, so I think AI is such a broad topic in the context of automotive because there is, of course, the driver assistance and automated driving aspect of AI. And then, there is the consumption of AI as a driver. And then finally, there is maybe the consumption of AI as an automaker, and what do you do? How do you build for a product that has a very long life cycle? Maybe to start off from the first one, we have seen, I think the automotive industry has actually embraced AI, going back 10 plus years with classical AI. And classical segmentation, and classification, and labeling of objects, and all of that to now adoption of transformer models and being able to look at a scene and process the scene, and being able to actually navigate it in a very natural way. If you look at the broad spectrum of solutions that are deployed today, that is everything from the classical systems that are driver assistance based to the fully automated robot taxis. And they are obviously using a very wide variety of AI technologies.

Our realization is the following, there is silicon available for being able to support every single dimension of AI that exists out there. We are investing in all of those, as are many of our competitors. So this is no longer an issue where your silicon limit. There is a tremendous amount of data available. So you can certainly train models for every different type of AI solution that is out there. So what it really comes down to is how quickly can you embrace AI and automated driving? Where you draw the line between what is safe and where you can trust the machine completely versus where you need to actually have guardrails, where you need to be able to bring in experiences that the auto industry has gained over a very long period of time. I think the other dimension is what is the variability in the type of environment you’re driving in when you’re driving in an environment where the infrastructure is advanced, where the infrastructure is designed for the roads to be applied with automated driving.

I think that’s one scenario where you are in ODDs, in operational driving domains that are less designed for that type of a situation. What type of complexity does that pose? So at the end of the day it really just comes down to which market embraces the technology faster. Part of it is consumer adoption, part of it is safety regulation, but I think we are now moving in this direction where you are going to be able to have more and more safety built in into cars natively, and the progression towards how does your commute become more convenient, how do you not have to deal with stop and go driving? I think those things will start to happen in a much more natural fashion. Where I think there will still be concern is high speeds when you are in a high speed environment where safety and the impact of being lax on safety much higher. How do you deal with a situation like that?

I think the other areas will obviously be in highly dense environments where there are lots of pedestrians, there are lots of mixed traffic scenarios. How do you deal with those types of safety type situations? Can you be in a fully automated environment very easily? So I think that it is a complex topic because it has a lot to do with risk assessment, and risk is assessed from the eyes of the automaker, and different automakers have different types of experiences depending upon how long they’ve been at it, which part of the world they have originated from, what kind of footprint they have in place. Our approach is multi-fold from a silicon perspective, we are building silicon that allows us to touch every single form of automated driving, from basic driver assistance to the most advanced. From a stack perspective, we are being careful.

We want to make sure that we are building solutions that hit the sweet spot that are designed to be safe. We are partnered with some of the most advanced companies when it comes to this, BMW is a key partner of ours. And we are taking those learnings, we are applying them, we are adopting those practices, and then we are engaging such that when we build the next generation solution, we are thinking about those things very holistically. One big shift that we have made in the company over the last several years is to take a safety mindset when it comes to automotive. We are a smartphone company at our roots, but this transition over the last several years has allowed us to actually make sure that we are thinking about quality and safety as core tenets to our overall strategy. And so I think it applies very heavily, especially when you talk about ADAS.

If you then think about AI outside of driver assistance in the car, I think the one thing that is really unique about the car environment is that it is much more dynamic than a smartphone or a PC. You are moving different parts of the world, different experiences, the journey from A to B gives you a lot of different contexts. And so the multi-modality of AI I think is highly visible, is highly at play in terms of how to take advantage of AI in a vehicle environment. So some examples, you have to be able to speak to your car in a natural way because it is easier than being able to press buttons and look for a touch screen. Now with the types of models that voice-to-text has enabled, I think that’s going to become fairly straightforward. I think that’s just going to get deployed everywhere, in some parts of the world, that is already the case.
Once you go do that, the whole conversational nature of AI, how do I actually get an answer to what I’m looking for with a tremendous amount of context? The car knows where I am, it knows what time of day it is, it likely knows where I’m headed. And so there is a lot of intelligence in the prompt that is generated for the question that you’re going to ask. And as AI becomes smarter, you’re going to be able to get much more value out of that entire engagement. And so that is a journey that has started, and I think we are seeing the very early days of it. But really fascinating is, in terms of everything that is possible. The other thing that we see that is really interesting is because the car literally sees so much about what is around it, there are cameras, there are temperature sensors, there are proximity sensors. That adds a tremendous amount of input into what the model needs to be aware of when the user is interacting with it.

So, for example, if I’m driving in downtown, I know the environment I’m in, I know the traffic that I’m in, I know what time of day it is, I know what is open, what is closed. Those things add tremendous context to the conversation, makes a lot more intelligent.

I think the other piece which I feel automakers are taking a lot of advantage of now is the ability to do onboard AI, AI in the vehicle. So, for example, being able to talk to your vehicle, if you bought a new EV and you’re looking for the instruction manual, you don’t have to go take it out of your booklet. You can ask a question, and the manual is already in the vehicle, it is indexed, it is available to be searched through your voice command. If additional information is needed. It knows how to get that from the cloud. So this whole combination of hybrid AI, I think, is going to start to get adopted very quickly in the vehicle. So I think there are so many used cases, such a rich surface area for used cases, and then, a combination of what you can do onboard versus in the cloud makes this, I think, super interesting.

Olivier Blanchard:
Yeah, and thanks for… I was going to ask you about your partnership with BMW, specifically about ADAS, so thanks for answering that. All right, so it looks like we only have about one minute left. So let me ask you a quick question and see if you can give us a quick answer. I’ve been trying to explain to people what a software-defined vehicle is, and for me, it’s simple, but it’s not simple for everyone. So can you clarify, in a way, in a brief way, the concept of software-defined vehicles, or SDVs? Are we talking Buzzword? Are we talking… What’s the importance of SDVs in automotive nowadays, and why people should not be only thinking about EVs but should also be thinking about in terms of SDVs?

Nakul Duggal:
Yeah, I think if you think about the design of an electric vehicle, for example, it’s a highly simplified design because you have a battery bank, you have an electric powertrain, you have the cabin, and then you have autonomous driving. And for an automaker to be able to create an architecture where you can manage each of these elements independent of the other becomes very important. And so at the simplest level, a software-defined vehicle is to be able to manage the specific aspects of the vehicle that can be isolated, that can be modularized, manage those independently. For example, I want to manage my user experience based upon who I am. If the car is a rental vehicle, I want to be able to have a very versatile, very flexible experience. How do I separate that from the fact that this is an EV and those aspects need to be separated?
Software-defined vehicles allow you to be able to have a software architecture that allow for modularization of whatever has to be managed in the user domain versus in the vehicle domain. That’s probably the simplest way to go describe it. It’s a new way of writing software for the car as opposed to the hundreds of ECUs that cars used to have, where a variety of different people have to come together and write their software. And we are on that journey. That is something that we have already decided to go down. So I think you will start to see a lot of that take place in vehicles going forward.

Olivier Blanchard:
I expect So. Well, thanks for that. I’m going to have to have you on again because I have so many more questions to ask you. I’m just kind of sad that we’ve already run out of time. This 15 minutes went by really quickly. So thank you again for joining. For those of you watching us, thank you for joining. Don’t go anywhere. We have plenty more Six Five content coming, so stick around and enjoy the rest of our broadcast.