Not long ago, Tesla announced Robotaxi in August to stir up the market. The industry also expects that this “pangasius,” which changed the traditional automobile industry, will once again cause drastic changes in the driverless car industry.

Unexpectedly, Tesla was “beheaded” by Baidu.

On Apollo Day 2024 on May 15, Baidu revealed that its Robotaxi business Radish Express will fully enter a profit period in 2025.

This means that, with the support of the big model, the driverless car business, which bears Baidu's high hopes, is reaching an important milestone in commercialization. In the past, the cost of driverless smart cars remained high due to technology and hardware, and the path to commercialization was frequently blocked.

Chen Zhuo, general manager of Baidu's autonomous driving business department, revealed that at the beginning of this year, with significant improvements in spatio-temporal coverage and operational efficiency, Radish Express's revenue increased 9 times, and losses were reduced by more than half. Starting May 15, the first batch of sixth-generation unmanned vehicles delivered will be put into use in Wuhan, and the deployment of 1,000 unmanned vehicles will be completed within the year.

Chen Zhuo said that as 1,000 sixth-generation unmanned vehicles are put into use one after another, Radish Express's revenue will grow faster and is already very close to the breaking point of break-even. “Our goal is that by the end of 2024, Radish Express will achieve balance of payments in Wuhan and fully enter a profit period in 2025. Radish Run will be the world's first commercially profitable autonomous driving travel service platform.”

Baidu waited ten years for this result.

Su Tan, general manager of the smart car business department of Baidu's intelligent driving business group, believes that AI in the past had certain recognition capabilities, but was unable to handle complex scenarios. However, the scenarios faced by smart cockpits are becoming more and more complex. Facing highly complex scenarios, large end-to-end models are more advantageous in solving problems.

At the press conference, Baidu unveiled the Apollo ADFM (Autonomous Driving Foundation Model), the world's first large model to support L4 autonomous driving. The model reconstructs autonomous driving with large model technology. It can balance the safety and generality of the technology, achieve safety more than 10 times higher than that of human drivers, and achieve complex scene coverage throughout the city level.

Radish Express also released the sixth-generation unmanned vehicle, applying the “Baidu Apollo ADFM Large Model+Hardware Product+Safety Architecture” solution.

Compared with the previous generation car, the total cost of the Radish Express sixth-generation unmanned vehicle dropped by 60%, and the price was only 200,000 yuan, setting a new industry record. At the same time, once the construction of the automated operation network for the Radish Express Driverless Vehicle is completed, its operating costs will also be reduced by 30%. With late-stage autonomous driving technology and continuous optimization of human cabin efficiency, service costs will be further reduced by 80%.

Wang Liang, chief R&D architect of Baidu's intelligent driving business group and chairman of IDG's technical committee, believes that with the reduction in chip costs, the overall vehicle cost of driverless smart cars is expected to drop further and reach the 100,000 yuan range.

Wall Street News also learned from Baidu that as of April 19, the total number of service orders for Radish Express had exceeded 6 million. In the first quarter of 2024, Radish Express provided about 826,000 services to the public, an increase of 25% over the previous year; the operating area of Radish Express in Wuhan has exceeded 3,000 square kilometers, and the service covered more than 7 million people.

The reduction in the cost side and the increase in the number of services and area are the secrets of Radish Express's ability to take the lead in making a profit in Wuhan. Moreover, this plan in Wuhan is replicable. Xu Baoqiang, general manager of Baidu's autonomous driving technology department, revealed that based on the big model of autonomous driving, Baidu already has the technical capabilities required for city-level autonomous driving travel services. Facing a new city, it only takes half a year to complete preparations for a fully autonomous travel service throughout the region. In the future, this verification time will be further shortened to enable autonomous driving where Baidu maps are available.

After more than a year of technology competition, how to implement and commercialize big models is becoming a common proposition for the entire industry. The wave of big models is raging, but how to successfully land and achieve profits is still a fog for players entering the game.

With Radish Run, Baidu found a viable path in the big model competition. This has undoubtedly dropped a bombshell for the entire industry. In it, it seems that we can see the possibility of big models reshaping industries and applications.

Over the past 11 years, Baidu has invested more than 140 billion dollars to develop “All in AI”, and it has finally achieved good results. On Baidu's earnings conference call on May 16, Li Yanhong also stressed that the company is shifting from focusing on the Internet to prioritizing artificial intelligence, and is actively promoting the restructuring of To-C and To-B businesses using the Wenxin Big Model.

As can be seen, Baidu has also made quite a bit of progress in using AI to restructure Baidu's business.

Financial reports show that in the first quarter of 2024, Baidu's total revenue reached 31.5 billion yuan, with net profit (non-GAAP) attributable to Baidu of 7.011 billion yuan, an increase of 22% over the previous year; Baidu's core revenue was 23.8 billion yuan, and net profit (non-GAAP) attributable to Baidu's core was 6.628 billion yuan, an increase of 26% over the previous year.

Specifically, Baidu Smart Cloud's revenue for the first quarter was 4.7 billion yuan, up 12% year on year, and continued to achieve profit (non-GAAP). Among them, the share of revenue contributed by generative AI reached 6.9%, and the share is gradually expanding.

Institutions, including Morgan Stanley, are also optimistic that AI will boost Baidu's business. For example, the revenue growth of cloud services, which has the most direct impact on AI, is expected to continue to grow in the 2024 fiscal year. Morgan Stanley said it is optimistic about Baidu's dominant position in China's AI field.

Li Yanhong also revealed at the earnings conference that following cooperation with China's Samsung and Honor in the previous quarter, mobile phone manufacturers such as Xiaomi, OPPO, and Vivo all connected to the Wenxin Big Model API this quarter. In addition, the field of cooperation for large models is also expanding from mobile phones to PCs and electric vehicles, and Lenovo and NIO Auto also reached cooperation during the quarter.

Meanwhile, Baidu's basic platform is also being reshaped by AI. Currently, 11% of search results on Baidu Search are generated by AI. Li Yanhong also said with full confidence, “Search is most likely to become a killer app in the AI era.”

In the past, Baidu was always one step ahead in investing in AI technology, and itself seemed like a “Huangpu Military Academy” for the driverless driving industry, but in the end, the pace of implementation and commercialization always seemed unsatisfactory. Now, Baidu has finally found some direction. Next, it is how to further sort out and clarify these explorations so that performance can be further improved. This will be a key part of Li Yanhong's transformation in the AI era.