Nvidia’s Danny Shapiro: "Our approach is fundamentally different to the competition"
Technology and Business
NVIDIA has quickly turned into one of the largest players in automated driving. In an exclusive interview, Danny Shapiro, Senior Director of Automotive, explains how to develop safe driverless cars – and how NVIDIA is growing its ecosystem with more than 300 automotive companies.
There are countless predictions regarding automated driving. Who will be the first carmaker to put self-driving vehicles on our roads? How many of those cars will be sold in ten years? The answers will vary, depending on what study you read. Which just goes to show: forecasting in this field is tough. But there is one prediction many experts would instantly agree on: whoever wins the race for driverless mobility will probably rely on NVIDIA technology. The AI computing company has secured a key role in the development of autonomous driving systems, and is now moving into production. We spoke to Danny Shapiro, Senior Director of Automotive, on challenges and responsibilities that arise from this unique market position.
2025AD: Mr. Shapiro, what was your first encounter with automated driving?
Danny Shapiro: My first experience inside an autonomous vehicle was at CES several years ago when BMW ran an automated car on a closed race track. It wasn’t truly autonomous, but it was a computer-controlled vehicle that drove a fixed route at high-speed, with nobody touching the wheel. It was a transformative experience. Since then I’ve ridden in many autonomous vehicles. I drive an automated car to work every day, powered by NVIDIA DRIVE. It’s a Level 2 system, which means I am responsible and still have to monitor it. However, having an AI computer handling routine driving tasks means it doesn’t bother to be in a traffic jam. My commute experience is much more relaxing.
2025AD: NVIDIA has been very successful in designing Graphic Processing Units (GPU) for video games. But your company turned into a huge player in the autonomous driving space within just a couple of years. How did this come about?
Shapiro: Carmakers have been using NVIDIA GPU technology to design, engineer and manufacture their cars for two decades, and it was about ten years ago that we started using our processors and software inside the vehicle to improve the user experience. Initially that meant bringing consumer electronics style graphics and interactivity into infotainment systems. Then as our mobile computing horsepower dramatically increased, we introduced AI technology to accelerate the development of automated vehicles. Our relationships with numerous vehicle makers has put us in pole position for supplying compute and artificial intelligence technology – far ahead of any other supplier.
2025AD: With automated driving, NVIDIA technology could become crucial for the safety of billions of drivers. How do you deal with this enormous responsibility?
Shapiro: Safety is our number one priority. If there is a problem with the navigation system, it’s not life critical. But as we move toward fully autonomous driving, safety is paramount. NVIDIA’s solutions adhere to the strictest safety requirements. Within the industry, there is a race to develop self-driving vehicles and put them on the road, however we need to ensure the safety of the systems. To that end, at CES earlier this year we announced the world’s first functionally safe AI self-driving system, which uses redundant and diverse functions to make sure a vehicle will operate safely, even in the event of faults related to the operator, environment or systems.
2025AD: How can you ensure the safety of the systems?
There is no way we can generate enough miles to train and validate our deep neural networks only relying on real-world driving. So, NVIDIA has developed solutions for simulation. Building on our company’s video gaming heritage, we can simulate all different driving scenarios far more efficiently. Will the car detect a pedestrian at night in a rain storm? Will it see that child running in front of a vehicle from behind a parked car? In a datacenter, using our deep-learning supercomputer known as DGX, we can simulate 60,000 miles of driving in just one hour. That means we could simulate driving every single road in the US in just two days. These simulated miles are used to not only train complex deep neural networks, but can help validate the safety of autonomous driving algorithms.
2025AD: You mentioned the race among all players towards fully autonomous driving – which could also be seen at this year’s CES which was overflowing with bold announcements. Are we experiencing a hype?
Shapiro: Some people call it hype, but I do not agree with that. It really is a transformative time. There is a lot of news being reported simply because there is a great deal of activity and advancements being made. Virtually every automaker is working on some form of automated driving. Of course, not every company is going to make it. It is true that we’re working with over 320 partners that develop autonomous driving solutions on our NVIDIA DRIVE platform. This includes over 150 startups working on new sensor technology, mapping or algorithms. Some of these companies might not still be around at the end of this year. Others might be bought by bigger companies. But the truth of the matter is that we will see great advancements this year and in the next few years to come.
2025AD: With NVIDIA having more than 320 partners, to what extent is it possible for carmakers to customize products to allow for differentiation?
Shapiro: Developing an autonomous vehicle computer is very different than supplying a component. We do not sell a fixed product for a carmaker to simply stick in their car. Rather, NVIDIA has a true development relationship with carmakers. We are building programmable computer systems comprised of hardware and a massive amount of software. NVIDIA DRIVE is an open system, meaning our customers build their applications on top of our technology. That gives them the ability to customize the user experience, the features of that vehicle, and the level of automation. They will also have the ability to update the car over time – much like updates for your smartphone. For instance, Tesla built their car around NVIDIA technology. With software updates, they can change the user interface, they can change driving dynamics, and they can add new autonomous capabilities.
2025AD: Your CEO Jensen Huang announced at CES the DRIVE Xavier, the new processor with an unheard-of amount of processing power, enables fully autonomous driving. Would you consider the question of providing sufficient computing power for driverless cars solved?
Shapiro: There is never enough computing performance. The deep neural networks that we are running will continue to become more and more complex. There will be more sensors on vehicles, generating more data – so the need for computation in the car will continue to increase. The critical issue here is safety. We will need to increase the level of redundancy in systems. DRIVE Xavier delivers the highest level of compute performance today, and its energy efficiency is designed to enable production AV systems. But for a level 5 Robotaxi, we envision multiple Xavier processors plus discrete GPUs being combined to deliver the computation required. Much like humans are constantly learning, these cars will continue to get smarter and smarter over time.
2025AD: One of your biggest competitors is Intel. Having acquired Mobileye in 2017, they claim that the approach of ever increasing the amount of data is flawed and that they aim to collect only as much data as necessary – to keep data sets and costs manageable. What do you say to that?
Shapiro: Our competition has a fundamentally different approach. We believe autonomous driving is a super-computing problem. The more data you have, the higher your level of reliability. The other approach is: being a component supplier, having a very low-energy, low-compute solution like smart cameras, and trying to use that to drive an autonomous vehicle. It just won’t work. If you look at all the companies that are developing level 5 robotaxis, essentially every single one of them is using NVIDIA. They recognize the need for massive amount compute cycles versus trying to leverage an array of smart camera processors.
2025AD: What are the hurdles that still lie ahead for NVIDIA on the road to fully automated cars?
Shapiro: We will continue to work on delivering vehicles that are remarkably safer than any human-driven vehicle. But as a society, we need to figure out: does an AV need to be two times safer? Ten times safer? A hundred times safer? At what point does society say: we are ready to put these vehicles on our roads? I don’t know whether it will be possible to prevent every accident out there as long as many cars are driven by humans. But we will dramatically reduce the number of accidents and fatalities.
About our expert:
Danny Shapiro is NVIDIA’s Senior Director of Automotive, focusing on artificial intelligence (AI) solutions for self-driving cars, trucks and shuttles. He's a 25-year veteran of the computer graphics and semiconductor industries, and has been with NVIDIA since 2009. Prior to NVIDIA, Danny served in marketing, business development and engineering roles at ATI, 3Dlabs, Silicon Graphics and Digital Equipment. He holds a BSE in electrical engineering and computer science from Princeton University and an MBA from the Hass School of Business at UC Berkeley.
Submit your story
Become part of our autonomous revolution and submit your stories, images and videos
Stay up to speed with our weekly briefing. Enjoy autonomous driving content direct to your inbox