Nvidia's Huang predicts software will rule industry

Nvidia CEO Jensen Huang expects the company he founded in 1993 to play a key role in future self-driving vehicles. To do it, he wants to put together the "brain" of autonomous-driving systems. Huang, 57, predicts it will be commonplace for automakers to sell new cars at cost within four years because profits will instead come from the manufacturer's proprietary software. He recently shared his vision in a video chat with Automotive News Europe Associate Publisher and Editor Luca Ciferri and Managing Editor Douglas A. Bolduc. Here are edited excerpts.

Q: How many fully autonomous vehicles will be on the road by 2030, and what percentage of them will have Nvidia technology?

A: By 2030, I would estimate that 20 percent of the cars on the road will have a high level of automation — and most of them will have Nvidia's technology inside. Also, the vehicles making up the 20 percent will account for 50 percent of the miles driven in the locations where they are available. And all of them will be able to work around the clock. They never get tired, they never lose their concentration. So their utilization rate is going to be much higher.

When you say a high level of automation, do you mean Level 4 or even Level 5, which means the vehicle no longer needs a driver?

We will see Level 5 for the delivery of goods on campuses, at compounds or other areas that can be closed off. It is the most cost effective and safest option because, unlike humans, robots don't mind going slow. Robots have plenty of time. They don't get impatient. Therefore, those areas should be completely Level 5.

What about for passenger cars?

By 2030, I think autonomous vehicles will be largely Level 2, but these Level 2 vehicles will be incredible. It will almost be like driving with your mind. You think it and the car does it all while keeping you out of harm's way. By then, parking and retrieving your car, highway driving and driving in traffic jams will be completely autonomous. Level 2 alone is going to be a completely transformative driving experience.

Artificial intelligence and robotics are not supposed to replace the need for humans. They will enhance our work. They won't replace driving, but they will enhance driving. And while I really enjoy driving, I also love using the Autopilot in my Tesla. It is very relaxing.

What are the biggest challenges facing the auto industry over the rest of this decade?

The industry is going through several revolutions at the same time. One is the move toward electrification. Many people thought EV buyers were motivated by environmental concerns, but that was wrong. Yes, EV buyers care about the environment. However, an electric car is much more about luxury. It's ride is silky smooth, it's quiet, it doesn't make any bad smells and its full every morning because you can charge at home. I haven't been to a gas station in years. That's a luxury. The industry initially missed this very important point, which is that this new functionality creates joy and delight through its convenience.

The second revolution is that the industry will have to change the way it develops products in the future.

What do you mean?

A tremendous amount of the user experience, capability and functionality in the car will come through software. This will start the moment you sit down, with the way that the car addresses you, engages with you and remembers your habits and preferences. The user experience is delightful because it almost seems like the car anticipates your needs whether you are driving or not.

How will that change the traditional automotive business model?

One of the most important things customers will value is the ever-progressing, ever-developing, ever-enhancing software. Since customers have the potential to be delighted by new software for as long as they own the car, the business model will fundamentally change. It is very likely that many car companies will sell cars at near cost and largely deliver benefits to you through the software that you subscribe to, or that you buy on a regular basis. When that happens and the car becomes software-defined, it will be imperative for the automotive industry to become excellent at software.

Why?

Because that's the only way to capture value again. We have seen this in many other industries and it is very likely we will see this in the automotive industry.

How will this transformation be different for automakers compared with other industries?

What separates the auto industry from the mobile phone and television industries is that the software that runs in the car is bespoke to the car. Therefore, the automaker will largely own the software opportunity, in addition to the car, for a very long time. That's fundamentally different than the mobile phone industry, where the mobile phone makers, with the exception of a few, largely don't benefit from the software on top.

What is the long-term effect?

In the past, the functionality of the car was complete at the point of sale. In the future, the functionality of the car is largely incomplete at the point of sale, but it needs to have a great capacity for software. That way the functionality can be continuously enhanced by car companies for as long as the customer owns the vehicle.

That creates tremendous economic opportunity. When you do the simple math of installed base times software opportunity, it is measured in hundreds of billions of dollars. This could very well be the world's largest single economy.

What is the profit potential?

Let me just do some simple math.

If a car company manufactures 10 million cars a year and sells them at cost with the potential to generate $5,000 of profit in software, that is $50 billion a year. Of course, you have to harvest that opportunity by writing great software, but it's still $50 billion a year of profit opportunity on top of making back the cost of the car. This profit completely belongs to the car company that built it.

Does the $5,000 come at once or every year throughout the life cycle of the model?

Many people will decide against upgrading the software at the time of purchase.

However, after two years they might be getting a little bit tired of the car. Then they find out that the car company has written a fantastic piece of software, and for $1,000, or maybe $5,000, they can make the car feel like it's new again. This upgrade also could be made by the customer before selling the car.

So, whether it's at the point of purchase or during the ownership of the car or to boost the car's resale value, there are so many opportunities to enhance that car.

How long will this transformation take? Until 2025?

Most car companies will have to make a major transition within four years. The reason is because I believe mainstream cars will be sold at cost. There will be no profit margin for mainstream cars. The profit margin, however, will be enormously rich in software.

Why four years?

Because in four years' time it will be commonplace for a very good car to be sold at cost for $25,000 to $30,000, and you can decide later whether you will buy some software. It will also take four years for the technology to be developed. This is the new business model. That's why the traditional mainstream car companies must make it happen in four years. It's imperative.

Which companies have an advantage?

It's easier for startup companies to do this because there is no legacy to protect. It's a very challenging transition, but the reason why it is so easy to capture the courage needed to do this.

There is a fear that this transition will result in a massive reduction in jobs. Do you agree?

The car industry will absolutely hire more people. It just has to get to the other side of this transformation because where there are great economics, there is great opportunity.

How will Nvidia and other technology-focused companies benefit from this change? Will they take over more control of the car?

We have changed the way we offer our technology so that we can enable car companies to create their own fleets and determine their own destiny. The idea that the technology industry would control everything inside the car makes no sense. It won't happen. The reason for that is very simple — the car companies are going to be fleet managers and service providers, not widget makers.

Will Nvidia be a hardware supplier, a software supplier or a supplier of both?

Nvidia is a full platform supplier that works with the car industry however the different companies choose. For example, robotaxi companies want to develop all of their own software and operate the service themselves, but they would like to purchase Nvidia's computing solutions, and use our AI ecosystem and tools. Some customers would like us to build a whole stack so they can have the capability by 2022. We are not a carmaker. We are a technology maker that wants to bring this world of autonomous driving and AI to the automotive industry. There are many different types of companies and they have different needs. We have to be flexible.

This also requires a long-term outlook, right?

We have the benefit of being quite patient. We have been working on autonomous vehicles for six years. This requires a company with a lot of determination, a lot of staying power, a lot of indigenous core technology that we can leverage. But we are here for the long haul.