The role of AI in automotive: Developing smart EVs | Ganesh Iyer

All right, before I start my presentation, raise your hand if you have heard of this acronym called NIO. Wow, 80%. I see some very familiar faces as well. So my name is Ganesh. I joined this company about seven and a half years ago. I spent a little over four years at Tesla prior to joining NIO. So who is NIO? We call ourselves a lifestyle brand. Our vision is to shape a joyful lifestyle for our users by leveraging the next generation technologies in AI, autonomous driving, and e-powertrain engineering technologies. We are a pretty young company, conceived in 2012 but officially launched in November of 2014. As I mentioned, we strive to shape a joyful lifestyle for our users and become the best user enterprise in the world. And how do we find the users? The smart EV is a core product that we manufacture, and it's the way to find the users, nurture them, and grow into a community. We believe a community-based business model is the next big disruption, especially in the context of the automotive industry.

Where are we located? This is a quick snapshot of where NIO is located. When I joined, the company was less than about 900 employees. As we speak, we are about 36,000 plus, predominantly engineers around the world. On the left-hand side, that's where I head up San Jose, which is the Silicon Valley or North American headquarters. On the right-hand side is where the company was born back in Shanghai and Beijing, and we expanded to many other cities. In Europe, about a couple of years ago we entered the European continent, starting with Norway, and last year we entered five other markets: Germany, Sweden, Denmark, and the Netherlands.

So this is the second generation vehicles that we have deployed, meaning delivered as of today. NT2 stands for the next, the second generation of NIO Technology platform. We built the platform and have deployed eight distinct smart EVs: smart, connected, premium electric, and of course autonomous vehicles. The first generation we mainly built three products: ES8, ES6, and EC6. We don't manufacture those cars anymore. The second generation is what the team has been focusing on. Very excited to report as of yesterday, nearly 400,000 premium SUVs have been delivered to our users throughout mainland China as well as in Europe.

As we all know, the automotive industry and smart EV industry is changing dramatically from mechatronics to electronics, and now it is completely digital, not even software, everything is digital. The vehicle has to be much smarter than it is today, so that's why we call it a smart plus EV. That is where the future is heading. The entire vehicle needs to be upgradeable, be it software, firmware, or even the hardware itself. At NIO, many of you are familiar with the term FOTA, firmware over-the-air update. Back in 2012, 2013, Tesla started that process, and at NIO now, nearly 100% of the entire software and firmware bits can be easily upgraded and pushed to our vehicle fleet. It's completely upgradeable. I'll touch on the hardware upgrade later. Essentially, the entire battery, which is the single most expensive component in any electric vehicle to date, be it Tesla or any other model, low end to premium luxury segment. Once you buy an expensive battery pack, unless it's upgradeable, it's a concern. We focused on how to solve that use case for users. I will touch on that later.

Horsepower is now changing to compute power. As the vehicle becomes much smarter, it's all about the computing capability that exists within the chipset or the associated systems. It's one of the critical factors. The vehicle is moving towards a lightweight approach, no longer steel-based chassis, mostly moving to lightweight aluminum and even carbon fiber to a certain extent. There's a lot of innovation going in that direction. I was in Europe three and a half weeks ago at an automotive OEM symposium where almost all OEMs are talking about no longer the vehicle or the product they sell, but mainly the user experience. Everything is moving towards a more immersive user experience-based selling. Otherwise, customers like me and you, there's no differentiation. It needs to be differentiated by unique user experiences. The whole industry is now moving towards autonomous, whether it's a small robot all the way through vehicles. Starting with Tesla, now pretty much every OEM is investing billions of dollars towards getting to that stage. I jokingly say even steering as an option, maybe a day will come where you may be asked, 'Would you like a steering wheel?' That's where the industry and innovation is heading.

Putting that into the context of NIO, I'll touch on a couple of AI applications that are very relevant for the smart EV industry. Battery is one area that still has more innovation to come. We started with lithium-ion batteries. NIO's predominant supplier is CATL. They give us the lithium-ion cells, and we have our own factory where we put those cells into modules, and the modules go into a pack. There's a lot of innovation in the packaging of the cells, which we are extremely proud of. The other aspect is the e-powertrain, a fundamental shift in innovation from traditional motors to more of an e-powertrain. We have a lot of innovations going on there, and it's owned by our proprietary factory in Nanjing, a city just outside of Shanghai. The last three are the digital brain of the car. From the top, it's a smart gateway. Most Tier 1 OEMs sell ECUs, electronic control units, which you can buy and integrate within your vehicle ecosystem. But the challenge with that is you are limited at the mercy of the Tier 1 or Tier 2 OEMs. If you are a startup like we are, only an eight and a half year old company still able to deliver nearly 400,000 units, how can we innovate faster without getting constrained by the Tier 1s or Tier 2s or any of those partners and suppliers? We created our own internal smart gateway starting with our first generation NT1 products, all those eight models you saw in NT2, and we are working on NT3 where one of the core innovation areas is mainly the smart gateway which sends the controls to all of the ECUs. It's the digital brain of the car.

Next is the smart cockpit or the digital cockpit. You'll see that little emoji-type thing on that image. We call her Nomi, spelled N-O-M-I. Nomi is the digital companion, NIO's digital companion. It's the first production car instantiation of AI. It's a conversational AI engine. Nomi is physical, as you can see, a little physical robot which sits on your dashboard. Nomi is contextual. Based on the context and facial recognition, it understands your emotions based on the context. Lastly, it's emotional. Depending on the situation or your feelings, Nomi also reacts accordingly, so you feel like you have a companion inside the car. For those who would like to see this, we have two vehicles parked in the visitor parking. You can see Nomi in action, play with her, and get a feel of it. That's the first AI application at NIO that we've deployed at scale. During lunchtime, somebody was asking about the take rate of that feature in the car. In the Norway market, it's more than 95% of the sales include Nomi as a feature because users who have little children find it pretty interactive, pretty cool, and fun to work with. You can control all the controls inside the car using Nomi.

Last but not least, autonomous driving, which is a complex area. We have a few individuals here, ex-NIO, who were instrumental in architecting the first generation of our autonomous driving ADAS features. For NT1, we partnered with Mobileye. In fact, NIO is the first OEM to work directly with Mobileye from his Royal team. We bought their perception stack and integrated it within our system. NT1 was all that. But the second generation, the eight models I showed you earlier, we partner with Nvidia. The Orin chipset, there are four of them inside the car. I will go through some of the specs later. That's a heavy computing platform inside the car. Even though all the cars have all the hardware required for level three plus or even close to level four, we haven't fully deployed the algorithms in the car. In fact, a couple of weeks ago, September 21st to be exact, we started pushing the level three plus algorithms inside the fleet across mainland China. The teams have done a tremendous job collecting millions and millions of miles of data to make sure these vehicles are safe and secure to be on the mainland streets of China. We are extremely proud of that. Now the vehicle is fully autonomous back in China. We are doing the same thing in Europe as we speak, and hopefully in the near future, we hope to enter the US market as well.

A little bit more in-depth on the Nomi AI application in smart EV. It's the world's first in-vehicle artificial intelligence, and it is contextual and conversational AI. This is slightly outdated. Now more than 2 million conversations are happening per day. People are interacting with Nomi, and nearly 2 million conversations are being processed interactively by Nomi. Let's take a look at some of the normal use cases that Nomi does.

So quick demonstration. I recorded this a couple of days ago from our vehicle lab in our headquarters here in San Jose. The inspiration for Nomi as a conversational AI engine in the context of automotive came mainly from our users. Our product management team went to the field and collected feedback from today's vehicle owners, vehicle drivers, soccer moms, baseball dads, students, doctors, etc. What we learned is not a secret. Most accidents are caused by human interaction as drivers. But all of us should be able to talk while driving. The inspiration for Nomi came from that data point from the field. We are extremely proud that now Nomi speaks multiple languages. We are teaching her even more languages. One day, hopefully every language around the world, she'll be able to focus and become your digital companion. It's not just the car controls. Our vision is to enable Nomi to be the entire ecosystem player. You should be able to connect to your home appliances and whatnot one day. We are working on that innovation to make her even smarter. That's one cool example of AI.

As I mentioned, it's not possible without a powerful computing platform behind it. We internally call that platform Adam, which is our super computing platform. It's mainly the NAD, the NIO Autonomous Driving backbone. We are leveraging Nvidia's Orin chipset, and it has a computing power of about 106 teraflops. Do we need it today? Probably not, but we are scaling this. These products are here for the future. We want to create as much computing as possible. It has a 48 CPU. You can read those specs. We are extremely proud of Adam, which is powering our second generation vehicle platforms and hopefully future generations.

The other cool AI application or robotics application within NIO. Before I go into the details, back in 2014 when we started the company, we thought about what is the main reason impeding the broader adoption of electric vehicles today around the world. It is concerns around the battery itself, the price of the battery. Those of you who have been with Tesla from the beginning can relate. When Tesla started selling the first Model S, the price of a battery pack used to be about $20,000, similar to the price of a Honda Civic. Granted, the price of the battery pack has come down dramatically, but still today, the single most expensive component is still the battery. We thought, how do we reduce that concern of the battery pricing so that we can put electric vehicles in the hands of middle class people who do not have as much purchasing power? We went to the drawing board and said we need to provide an end-to-end power replenishment solution, no matter which segment or profile of the user. How can we give electric vehicles in the hands of everyone? Internally, we came up with this: chargeable, swappable, and upgradable power replenishment. Chargeable is nothing but putting home chargers in your dwellings, apartments or whatever. We provide small slow chargers as well as fast chargers. But that's still not good enough for the charging experience. You want the speed of sound one day. The most cool robotics application we came up with was the swappable, which is one of our crown jewels. Back in the early days of Tesla, they tried to do swapping, but to me that's not required for countries like the United States and Europe where population density is relatively lower compared to China, India, or others where you don't have the luxury of real estate to put supercharger networks in every corner. How can we solve the problem? We thought about having battery swapping as one of the key innovations.

Though it looks complex, it's as simple as that. It's literally the size of this stage: three parking spots. One parking spot is for the incoming car coming for a swap, the other two parking spots for the robotic systems. You just drive in. We are on the third generation of NIO Power Swap stations. Essentially, you drive your car, through a proprietary app you can say, 'I want to schedule a swap at the Sacramento area power swap station.' Once the car reaches there, you park your car in front of the swap station, push a button on your instrument cluster, and the car will reverse and park itself into that empty spot. Then you have four minutes to decide whether you want to get out of the car or get locked in for four minutes. The robotic system comes in, slightly lifts the car, unscrews the battery, removes it, puts it into an empty slot, takes a fully charged battery, puts it back in, screws it back, and does a validation. You're off. It sounds simple, but as you can imagine, smart engineers in the room know it's a complex piece of engineering. We believe this is going to help drive more electric vehicle adoption because the fastest supercharger out there still takes about 20 minutes to get nearly 80% state of charge. That's still a lot of time. As the fleet size increases, there could be queues. We thought, today the premium for the early adopters and innovators is time. How can we give the time back? With the swap station innovation, you should be able to get your entire vehicle battery swapped in less than 4 minutes, giving that extra 15 minutes back for doing something productive.

This innovation also helped us do one last thing: the upgradable. You don't have to own or purchase a 100 kWh battery, for example, because it comes with a price. With the swap station and swapping technology being proven, you can go with the entry level battery pack, let's say a 70 kWh battery, for your normal commute. But you can swap out your battery as many times as you want, so you never have range anxiety. You may still have one or two use cases a year where you really need that extra range. But if you want to upgrade your vehicle, that's where the hardware upgrade part comes in. Go to one of the swap stations, pay a nominal upgrade fee, and you can upgrade your previously owned 70 kWh battery with a 100 or even 150 kWh battery in the future. That gives you the flexibility of not only upgrading your software but also the entire battery itself. This is another very cool AI application, robotics application that is driving more smart EVs to the mainstream.

These are some stats. Right now, it is nearly 29.5 million swaps. I just looked at the stats before I came here. Every minute and 20 seconds, a swap is being performed. The technology is proven. With that technology being proven, we said, why can't we come up with a new business model? Because the most expensive component in the EV is the battery itself. Why can't we give more purchasing power to those who couldn't afford it? Now that the technology is proven, why can't we break the smart EV into two financial transactions: one without the vehicle, which is the chassis plus all the power electronics and EDS, and then you pay a monthly subscription fee for the battery? The whole world is moving to an as-a-service model. We can reduce charging anxiety, potentially your residual value going down of your investment for electric, all that. Leave it to us. You just subscribe to a 70 kWh battery. It reduces the initial purchase price of the car. You don't have to worry about the residual value going down of the battery, and the safety aspects. As our engineers innovate new battery cell chemistry or even move to solid state in the future, you get the latest technology enabled battery cells and pack in your car. This is where we think the industry needs to move in order to drive more EV adoption. We have proven it now. We are extremely proud that over nearly 1,500 approved patents, still counting. The third generation swap station is capable of holding about 21 batteries fully charged. This is also bidirectional, meaning if the swap station is fully underutilized, it should be able to give energy to the grid and draw power. That will be another potential use case in the future. Tesla came up with the Powerwall, which mainly powers a small home, but it doesn't have as much capacity. This could be another use case for drawing and giving energy to the grid during peak time if you choose to.

This is the second real use case application in robotics and AI. Since the vehicles are now connected, before the patient comes to the swap station for swapping, we can proactively diagnose and get the health of each and every car out in the field. Right now it's 400,000. We know exactly what the health is. If the incoming car has a slight safety concern or whatever concern, we can actually swap, take the battery out of rotation, and do a detailed disposition plan to figure out before the battery can be put back and be safe again. This technology enables us to do all those normal use cases.

So this is our beautiful Innovation Center here in San Jose. We just moved in a month ago. This is our second generation office space here. Extremely proud. With that, thank you so very much. Happy to take any questions. I'll be outside during the break. Would love to chat and learn more about what other innovations we can do together to make this EV revolution a real thing at scale. Thank you.