Nvidia had a strong presence in automotive at CES 2024, focusing on generative AI and software-defined computing. They partnered with Li Auto, Volvo, Polestar, Mercedes, and other brands. Nvidia’s hardware for cars is increasing in performance; combined with Li Auto’s power management chip, the combined package claims 1,000 TOPS. They differentiate themselves through supercomputing expertise and programmability. Nvidia says its partnership with Mercedes involves using its processors in cars and digitalization efforts with Nvidia Omniverse and generative AI. Security is a priority, with measures in place to protect against hacking. Nvidia’s partnership with MediaTek brings Nvidia to different auto segments. At CES 2024, Nvidia’s partners showcased products powered by Nvidia technology, including autonomous trucks, vehicle computing solutions, and robotaxis.
What do we think? Nvidia is full of confidence, and there is no doubt the vehicles that their partners showed at CES were very lovely. We have previously seen the auto market prefer mobile-like GPUs with low thermal output and acceptable performance. The digitization of the cockpit continues at a fast pace, and in-vehicle infotainment (IVI) is becoming more important as EVs and self-driving trends develop. That, plus GenAI, is going to give Nvidia more opportunity with its higher-end solutions. And where they can’t deliver powerful, expensive solutions, they will use MediaTek to mop up the lower end.
One-on-one with Nvidia at the Mercedes booth during CES 2024
Nvidia had a strong partner lineup at CES 2024, and messages focused on the adoption of generative AI, along with the growing importance of software-defined computing.
We met with Danny Shapiro, Nvidia vice president of Automotive, to discuss some of the announcements and key trends, including digital twins, Omniverse-built factories, generative AI, and car configurators, along with Nvidia partners such as MediaTek.
Notable among the partners was Li Auto, a pioneer in extended-range electric vehicles (EVs), which has selected the Nvidia Drive Thor centralized car computer to power its next-generation fleets.
What does this next generation mean from the performance point of hardware that you can sell into a car. Where are we at?
Today, many automakers are using our Drive Orin system on a chip, which is 254 TOPS [trillion operations per second]. Li Auto is one of the first automakers to announce they will be using our next-gen processor, Drive Thor, which is 1,000 TOPS. And these innovative automakers will sometimes have two or more of these processors in the car.
The amount of compute you have in the car directly correlates with safety and what can be supported on the sensor suite, achieving higher resolution and higher frame rates, more diverse algorithms, deep neural networks, etc. Accelerated compute, like that of Drive Orin and eventually Drive Thor, is necessary to support the complexity of future AI models that will be implemented in-vehicle.
And there’s no problem shipping that level of performance in China?
There are a number of parameters the government has set over the last year regarding export controls. Everything we do complies with all regulations, and these new controls place restrictions on some of our data center products and highest-level gaming GPUs, but it does not apply to our automotive products. There is nothing in the vehicle that would cause concern with respect to national security, but rather, we are developing technology that is available globally to improve safety on our roads through automated driving.
In addition to passenger vehicles, we also work with robotaxis, trucks, shuttle companies, and delivery vehicles––from Mercedes-Benz, Volvo Cars, Polestar, Jaguar Land Rover to a range of Chinese EV brands like Li Auto, BYD, Xpeng, Nio, Zeekr. They are all integrating an Nvidia supercomputer into the car and then rolling out software updates to deliver various levels of automation, greater safety, and a future of upgraded functionality throughout the life of the vehicle.
And we know there will be new AIs discovered over this year that everyone’s going to want in their car next year. So, a carmaker needs to build in performance headroom and flexibility. That’s why innovative vehicle makers keep coming to us, asking for greater performance. You can never have enough compute.
How do you feel about your differentiation from all those sort of traditional Tier 1s, such as Renesas, Socionext, NXP, etc. Why are people shifting to Nvidia?
We’re about accelerated computing and AI. We’re about programmability. Our path in automotive was taking high-performance computing and bringing it to the car. We’re not starting with a low-end ECU and trying to turn it into an AI processor. Instead, we create high-performance, energy-efficient platforms that are programmable and open. The amount of software we develop is enormous, from the OS, programming tools, libraries to middleware, compilers, etc. And then, layer on simulation to test and validate the system.
EV makers like GWM (Great Wall Motor), Zeekr, and Xiaomi have adopted the Nvidia Drive Orin platform to power their intelligent automated driving systems. We are meeting at the Mercedes-Benz booth today. Tell me about these partners.
There are multiple ways in which we work with the automakers. One is what we’re doing with Mercedes-Benz. We are true development partners, architecting an updatable automated driving computer based on Drive Orin, using Drive OS, Nvidia software, middleware, and algorithms, with contributions from both engineering teams. In other cases, automakers may develop their own software stack directly on our platform.
But beyond automated and autonomous driving, we’re also working with the auto industry across the vehicle life cycle, from concept and styling, [to] design engineering, software development, smart factories, and retail.
This entire workflow digitalization is made possible with Nvidia Omniverse, our platform for integrating 3D applications to design, collaborate, plan, and operate systems that can improve the efficiency of the entire automotive industry.
Then we can layer generative AI on top of that. We are working with the design studios and introducing them to generative AI for design, concept, and styling. Think about designing interiors and exteriors from text prompts or sketches using generative AI. These are the kinds of really cool tools that are being developed. And then from there, we go into the engineering side, where we are unifying different departments. Instead of each team being siloed with their own applications, exporting and reimporting data where information can get lost, our platform is unifying designers, engineers, factory planners, and marketing teams.
On the topic of manufacturing, we are working closely with automakers on the factory side, helping redefine how the production line is laid out. In Omniverse, a complete digital twin of the factory can be created and run in simulation. This allows issues to be resolved and the entire factory tested––before construction starts. The time and resource savings are enormous. Mercedes-Benz is one of our customers using digital twins for production.
And they are also announcing new software-driven features, and the latest developments in Mercedes-Benz MB.OS, including the Concept CLA Class, which is using Nvidia Drive Orin for the automated driving domain. This is the first vehicle based on the next generation, Mercedes Modular Architecture, built on Nvidia Drive Orin, with additional models to follow thereafter.
The partnership between Nvidia and Mercedes-Benz is very different from the traditional supplier relationship. It’s not simply transactional, but rather, a long-term development partnership with revenue sharing between Mercedes-Benz and Nvidia. We both invest heavily, and both sides benefit from the success.
That’s exciting for Mercedes-Benz. Maybe not so good for consumers?
Actually, it’s great for consumers.
There are a variety of ways that vehicles can be configured, and there is flexibility with a software-defined approach. Most cars purchased today have a base price and then bundled features. You must decide at the point of sale what you want. With a software-defined car, it can be more customizable. You can select options at the time of sale, or later. The challenge, of course, is that when you’re paying for something physical, that’s one thing. But when it’s already in the car and then you must pay to unlock it, that’s a different mindset.
I think there’s a real worry that people will attempt to hack those systems, and then you start to make the whole car insecure because of that.
I hear what you’re saying. I think the reality is there’s been heightened alert on security for over a decade of cars.
We’ve worked to implement proven cybersecurity systems at the processor level, the board level, and the software. There are numerous security layers and software we pull in from our data center business: encryption, authentication, and virtualization. There’s a firewall to separate connectivity and infotainment systems from the safety-critical systems. We have put multiple safeguards in place, and we can even use AI anomaly detection, so if there’s something that isn’t part of the data flow in the vehicle, it can be stopped.
I’m very interested in the return of IVI. You know, it feels like everybody was shifting the conversation toward ADAS, that’s what matters. But now, with panoramic screens, for example, and the sort of generative AI features Sony/Honda showed, IVI is back to the forefront, with a focus on user creativity being reflected in it.
Yes, I believe AI is key to this. That’s where the next evolution of large language models is going, and a big part of that is being able to have brands customize the content and put guardrails in place so that the AI is an expert on everything to do with the car but doesn’t need to comment on other unrelated domains.
So, I guess this brings me right to my last question. You have this partnership with MediaTek. Is that a one-off, or would you consider similar deals with all these other incumbents in the Tier 1 space?
We have no exclusivity.
We’re bringing the AI processing down into segments we generally don’t compete in, at the entry level. Nvidia is focused on delivering the highest performance and premium experiences. We create advanced technology, and we have partners to bring it to additional markets that are complementary. The benefit of MediaTek automotive solutions is that it’s based on Nvidia GPU chiplets and our Drive OS software, coupled with the connectivity and graphics expertise of MediaTek. An automaker can select MediaTek solutions for entry level and Nvidia Drive for their automated and autonomous fleets, and the systems are supported with a single, unified software stack.
Writer’s addendum
As well as Mercedes, other key Nvidia partners at CES included:
- Kodiak, exhibiting an autonomous truck, which relies on Nvidia GPUs for high-performance compute to process the enormous quantities of data it collects from its cameras, radar, and lidar sensors.
- Lenovo, displaying its vehicle computing road map, featuring new products based on Nvidia Drive Thor, including: Lenovo XH1, a central compute unit for advanced driver-assistance systems and smart cockpit; Lenovo AH1, a Level 2++ ADAS domain controller unit; and Lenovo AD1, a Level 4 autonomous driving domain controller unit.
- Plus, showcasing its PlusDrive supervised autonomy solution on a highly automated truck, which uses Nvidia GPUs to process the trillions of operations per second its autonomous driving system requires.
- Polestar, showcasing Polestar 3, which is powered by the Nvidia Drive Orin central core computer.
- Zoox, showcasing the latest generation of its purpose-built robotaxi, which leverages Nvidia technology.