Baidu Guo Yang: Autonomous driving has reached a large-scale mass production node

In the fierce competition for smart cars in the Chinese market, new entrants still have the hope of catching up later.

During the Shanghai Auto Show, Guo Yang, chairman of baidu's intelligent driving business group product committee and general manager of the intelligent automobile business unit, made the above judgment when interviewed by the first financial reporter. In the Chinese market, Baidu is one of the first technology companies to invest in autonomous driving research and development. Guo Yang shared with reporters the latest trends in the development of autonomous driving software and hardware, he believes that China's intelligent car development trend exceeds other markets in the world, automatic driving technology has reached a large-scale mass production node, and "intelligence is such a vast market, now say who has occupied the market, I think this is impossible." ”

CBN: What are the main hardware costs of autonomous driving?

Guo Yang: The hardware cost of autonomous driving, first of all, is the sensor. Sensors include common cameras, then relatively inexpensive ultrasonic radar, and slightly more expensive millimeter-wave radar; and then recently hotter lidar. It is still difficult to assess the cost of lidar, because the magnitude of use is too small, and the cost of lidar must be very high in the case of only a few hundred vehicles and thousands of vehicles installed. If the installed capacity reaches hundreds of thousands of vehicles, millions of vehicles, or even tens of millions of vehicles, its cost may drop significantly. In addition to the cost of sensors, part of the cost of autonomous driving is the computing unit. With the improvement of automobile intelligence, the demand for AI algorithms, CPU logic algorithms, etc. will become higher and higher, and the cost of this part is relatively expensive.

CBN: What are the trends in the cost of software for autonomous driving?

Guo Yang: The early functions of the software are increasing, and the performance requirements are constantly improving, so the investment in research and development is increasing. The cost of software, if focused on a single function, is a downward trend in the future; but from the perspective of the entire smart car, the overall software cost will continue to increase in the next few years.

The cost of smart car software accounts for a fairly high proportion. Whether it is Tesla or Xiaopeng Motors, the proportion of software costs in the price is very high, and the cost of smart car software in the future will be much higher than that of hardware.

CBN: So is the iteration of autonomous driving technology accelerating or slowing down?

Guo Yang: It must be accelerated. Whether it is assisted driving in passenger cars or high-level Robotaxi, driverless technology is improving. To give two examples, last year Waymo directly launched driverless taxis in Phoenix, and Baidu also got 5 driverless licenses issued by the Beijing Municipal Government at the end of last year.

So, I think the trend of technology acceleration is very obvious, and a lot of new players joining has greatly accelerated this trend. So many new entrants want to make smart cars, which proves that the technology has reached a stage where it can do large-scale mass production.

CBN: What are the difficulties in mass production of smart cars?

Guo Yang: Mass production is not as easy as imagined. Mass production is very different from making something in a research lab. Mass production not only requires hardware to reach the vehicle specification level, the entire software system requires a lot of accumulation and testing.

Now the vehicle research and development cycle has been shortened, and many radical car companies used to have 3 years and 4 years to produce a generation of cars, and now it is shortened to one and a half years, two years to produce a new generation of cars. From the perspective of mass production, this is actually a relatively low cycle. If someone wants to shorten the cycle again, it still seems to violate the objective law.

CBN: At present, what are the commercial returns of Baidu's autonomous driving?

Guo Yang: We now have two models of cooperation with car manufacturers.

One is the traditional business model like Bosch and Delphi, which is to charge development fees, or collect some bicycle shares. But such a model is not in line with Baidu's current situation, because the products provided by Baidu are not the same as those provided by these traditional suppliers. Baidu now signs a service fee with the car factory at the delivery level: Baidu provides 3 years of service and 5 years of service to the car factory, and gradually adds new functions in these 3 to 5 years to achieve performance improvement.

The second business return comes from real user use. Many of the car companies that Baidu now cooperates with will adopt a subscription system, that is, Baidu hardware is embedded in it, and then all software will be subscribed to with the manufacturer. Users can take a one-time subscription, or they can use an annual or monthly subscription. Our current cooperation with many manufacturers will follow this model.

CBN: The development of autonomous driving to today, there have been a number of companies have made a lot of investment, walking in the front. Does this mean that latecomers have no chance?

Guo Yang: Judging from the current situation in China, I don't think so. When the first batch of new car-making forces collapsed two years ago, everyone felt that it was almost over, and the remaining few were all that the future had. But as you can see, nearly 10 new companies have come in this year. China is growing faster than other markets around the world because the environment is too competitive. Latecomers may have a certain latecomer advantage, but also have a certain time window loss. The market has just risen, intelligent such a vast market, now say who has occupied the market, I think this is impossible.