Mention autonomous driving in China and the first name that comes up will typically be Baidu. After all, it was the first Chinese tech company to put serious money into researching and developing autonomous driving technologies.

In late 2017, Baidu announced a RMB 10 billion ($1.5 billion) Apollo Fund to invest in 100 self-driving projects over the course of the next three years. The fund was the largest of its kind in the global industry.

Baidu also boasts a massive ecosystem of varied partnerships with more than 150 OEMs, Tier 1 suppliers, chip makers, and mobility firms. Meanwhile, just like its counterparts in the US, the Chinese search giant remains a highly-rated prospect as it vows to launch the country’s first robotaxi service by the end of this year.

But Baidu has fallen short. The company has been criticized for its slow progress in China, stumbling partnerships, and unfulfilled production plans. All of this culminated in rumors earlier this year that the company will spin off its self-driving unit after reporting its worst financial results in almost 15 years.

Drive I/O

Drive I/O is TechNode’s monthly newsletter on the cutting edge of mobility: EVs, AVs, and the companies trying to build them. Available to TechNode Squared subscribers.

What has Baidu actually achieved in its self-driving campaign? Does it really deserve the title of China’s AV leader? If the spin-off occurs, what kind of impact will it have on the company itself as well as the industry?

Baidu: the true leader?

Baidu began working on autonomous driving in 2013. Two years later, the company established its autonomous driving unit.

The company suffered a prolonged brain drain over the next 24 months, including the departure of Andrew Ng, an experienced AI expert and Stanford University professor. It was not until early 2017 that the company established its Intelligent Driving Group (IDG).

The company suffered a prolonged brain drain over the next 24 months. Baidu’s then deputy director Yu Kai quit in mid-2015, leaving to establish AI startup Horizon Robotics. In late 2016, James Peng and Lou Tiancheng, two of Baidu US scientists resigned and together formed AV firm Pony.ai in Silicon Valley. Tong Xianqiao, Heng Liang, and Zhou Guang followed suit one years later to found Roadstar.ai.

Baidu’s talent drain culminated in early 2017 when senior vice president Wang Jin left. He had helped Baidu form the AV unit from scratch. Finally, in March 2017 AI expert and Stanford University professor Andrew Ng resigned as chief scientist to form AV startup Drive.ai.
It was not until early 2017 that the company established its Intelligent Driving Group (IDG), led by Lu Qi, then president and chief operating officer of Baidu. The executive’s role in the newly minted group highlighted its strategically important position and core competency for the company.

The company seemed to be moving at light speed, kicking off Project Apollo in April 2017. Taking its name from the NASA’s missions to the moon, the initiative aimed to build a full-stack software and hardware platform for autonomous vehicles.

Three months later, Baidu made the decision to open-source Apollo—just as Google had done with its Android smartphone operating system—as it set its sights on catching up with its international counterparts.

Since then, Baidu has continued to release major updates every few months, building all aspects of driverless capabilities such as sensing, localization, perception, planning, and control—with tools including cloud services and open-source code.

When it released Version 3.5 earlier this year, Baidu maintained that the platform could handle the challenges of urban roads, including narrow lanes, speed bumps, and crossroads.

The company further ramped up its efforts to support mass production of driverless cars with the release of Apollo 5.0 in July, as well as an updated Apollo Enterprise, a suite of tailor-made autonomous driving solutions for OEMs that focused on robotaxis, minibuses, and valet parking.

In mid-2018, along with Chinese auto manufacturer King Long, Baidu began mass production of a driverless minibus called Apolong. The company also intends to launch a fleet of 100 robocabs in Changsha by the end of this year.

Nonetheless, these production plans have been far from successful. Baidu claims to have transported 40,000 passengers on its 100 Apolongs scattered around 20 Chinese cities. However, the minibus runs 10 km/h, completing an 800-meter journey in around 10 minutes. The vehicles are mostly used on public parks and high-tech campuses.

Baidu has also been late to running a robotaxi service. For the past nine months, the Chinese AV startup Pony.ai has been testing a fleet of dozens of cars, offering over 12,000 rides in a suburban area of 800 square kilometers in the southern city of Guangzhou.

Just as most global self-driving pioneers are backing off their commercialization plans, so too is Baidu. The company has increased its focus on more realistic goals: the connected vehicle market.

DuerOS, Baidu’s proprietary voice assistant, is among the platforms giving the company a competitive edge in the battle for Chinese drivers’ attention. Given the congested nature of China’s urban roads, onboard network services are gaining popularity. In June, Baidu announced it had partnered with more than 60 OEMs to install DuerOS for Apollo (the company’s voice-enabled vehicle connectivity platform) in around 300 car models, including Ford’s Edge ST SUV and Great Wall Motors’ top-selling Haval H6.

Chinese market research firm Gasgoo estimates that Baidu has overtaken former market leader Alibaba to become the leader in China’s vehicle operating system market. The search giant’s vehicle OS will be installed in more than 1 million cars by next year, according to Gasgoo, almost double the market share of Alibaba, which has been in a rocky tie-up with SAIC since 2018.

Baidu’s business with automakers does not look much better. The company had previously planned to work with state-owned BAIC and JAC Motors to produce Level 3 autonomous vehicles this year. However, sales of its Tesla-style enhanced driver assistance system Apollo Pilot were reportedly underwhelming, and the company has since shifted its focus to automated valet parking. Baidu denied the claims, but didn’t reveal further details.

Despite the company’s claims of cooperating with an extensive network of nearly 160 OEMs and key suppliers, a substantial number of them work nominally with Baidu. Instead of adopting Baidu’s in-car OS, automakers are instead opting to build their own software based on Android or Linux.

Automakers are willing to use some of the features the search engine giant provides, such as speech recognition and even its ad service, but are reticent about sharing data with the search giant. These companies see tech giants, including Baidu and Alibaba, as a threat and aim to make their core businesses untouchable, according to Wang Yao, a director at China Association of Automobile Manufacturers (CAAM).

Baidu’s Apollo project can lay claim to having done some pioneering work to push the industry. The company has gathered over 400,000 lines of code and 12,000 Github contributors; what’s more, it holds more than half of the road-testing licenses granted by Chinese authorities nationwide.

In an updated leaderboard released by Navigant Research in March, Baidu was placed in the category of “contenders,” chasing the big three leaders—Waymo, GM Cruise, and Ford—along with Toyota and Volkswagen.

However, the company has faced questions over its leadership in terms of commercialization and technological supremacy in the industry, with its position being challenged by Pony.ai. Baidu faces a rough road ahead.

The spin-off

Once the darling of China’s nascent tech industry, search giant Baidu is now struggling to keep up with its rivals.

Facing intensifying competition for advertising revenue and stricter regulation governing content as well as a sharp decline in the public’s trust and the spectre of the US-China trade war, the company’s future is clouded by uncertainty. Investors have certainly noticed: Baidu’s share price has fallen more than a third to around $100 since the beginning of the year.

As the company’s lead in search and advertising narrows, diversification has become ever more important for Baidu, with the company putting great emphasis on its self-driving and artificial intelligence initiatives.

But the success of Baidu’s autonomous driving program comes at great cost, which may ultimately be hurting the company’s bottom line.

“The diversification of Baidu’s business from mobile internet to the smart home, smart transportation, cloud, and autonomous driving markets will require heavy investments,” Baidu CFO Herman Yu said in the company’s 2018 year-end results.

In May, Baidu reported its first loss since going public in 2005. Shortly afterwards, rumors began to proliferate about an impending spin-off of its self-driving unit.

Autonomous driving spin-offs, real and rumored, have been big news in China this year. As the effects of the country’s capital winter continue to take their toll, companies are looking to independently finance their autonomous driving units.

Baidu may not be an exception. The company took a hit in the first quarter, reporting a loss of around RMB 330 million (around $46 million)—its first since listing in 2005.

Company CEO Robin Li had previously said Baidu would spin off the unit once it was mature. However, a company spokesperson said earlier this year that it has no such plans, adding that Apollo is an important part of the company’s AI strategy.

Baidu, which has been named one of China’s five AI champions—alongside companies like Alibaba, Tencent, and Sensetime—has its self-driving cars in numerous cities in China, including 45 in Beijing and another 100 expected to be deployed in Changsha.

The costs of getting these vehicles on the road are significant, TechNode contributor and co-founder of China Money Network Nina Xiang wrote recently, with each vehicle costing up to RMB 2 million. The cars in Beijing and Changsha may cost up to RMB 300 million collectively.

The company does not break down its R&D spending by business group, but Baidu’s cost of research has increased by 40% over the past 18 months to reach RMB 4.7 billion.

Vehicles, coupled with the cost of hiring engineers, particularly those in the US, are compounding Baidu’s financial burdens. The company has repeatedly attributed rising R&D costs to personnel-related spending. Baidu has between 1,500 and 2,000 employees working on Apollo.

Meanwhile, Baidu’s profit has fallen dramatically in the past year, plummeting to around RMB 2 billion in the first half of 2019 from more than RMB 13 billion during the same period a year earlier.

A spin-off is necessary for a company like Baidu, says Tu Le, the founder of Sino Auto Insights. He added that the return on current research and development costs won’t be reflected in the company’s books for the next ten years.

While talk of spin-offs increases, investment in autonomous vehicles has stagnated, making money harder to come by. The number of investments in Chinese AV companies peaked at almost 100 last year, but saw a sharp decline in the first half of 2019, according to figures from ITJuzi.

Investors are beginning to look beyond the initial overconfidence of the AV industry, in which companies promised highly autonomous vehicles in a matter of years. Massive investments at sky-high valuations are unlikely, especially for a Baidu spin-off when there are startups with better technology and longer-running robotaxi schemes looking for cash.

Jill Shen

Jill Shen is Shanghai-based technology reporter. She covers Chinese mobility, autonomous vehicles, and electric cars. Connect with her via e-mail: jill.shen@technode.com or Twitter: @yushan_shen

Chris Udemans

Christopher Udemans is TechNode's former Shanghai-based data and graphics reporter. He covered Chinese artificial intelligence, mobility, cleantech, and cybersecurity.