This article is authorized to be translated from the original article of Jiemian. Original title: 小米造车，组了个什么局? Author: Cheng Di.
Lei Jun, Co-Founder and CEO of Xiaomi, on Wednesday announced via his Weibo account that Xiaomi Auto Co. Ltd has been officially established, with registered capital of 10 billion yuan ($1.5 billion). Lei Jun himself serves as the legal representative of the company.
Since the official announcement that it would be entering the vehicle industry on March 30 this year, Xiaomi has done a lot of work.
For a start, the company has applied for many trademarks and patents related to automobiles. Lei Jun visited many car companies such as Changan Automobile, SAIC, Dongfeng, Trumpchi, SGMW and Great Wall Motors, as well as visiting the headquarters of Bosch and CATL.
Xiaomi Auto is to be headquartered in the Beijing Economical-Technological Area. The structural layout of car-building ambitions has gradually become clear. As far as the current situation is concerned, Xiaomi will definitely participate deeply in automatic driving technology.
Wang Xiang, Vice President of Xiaomi, also said that the first task of Xiaomi Auto at this stage is to find more than 500 talents focusing on L4 automatic driving technology. At present, Xiaomi has received more than 20,000 applicant resumes.
In addition to vigorously scoping out talent, Xiaomi is also constantly investing in and acquiring technology companies related to car-building, and is setting up various automobile teams.
Xiaomi officially announced that it will be involved in automobile field almost half a year ago. Now, let’s take a look the smartphone maker’s layout in the crowded field.
AI Interactive Ecological Chain: Xiao Ai Classmates × Daniel Povey
As early as April this year, many media outlets and experts speculated that Xiaomi Auto, in a way, did not have much of a chance to achieve a technological breakthrough that would shock the whole industry. However, it can indeed form its own development methods and characteristics, which is different from that launched by other new energy vehicle makers. The difference depends on Xiaomi‘s relatively closed ecological chain, that is, the artificial intelligence voice interaction engine “Xiao Ai Classmate” that consumers are very familiar with.
According to public information, the number of monthly active users of Xiao Ai Classmate was exceeded 100 million in 2020. The service is an indispensable part of various devices in Xiaomi‘s ecological chain, such as mobile phones, AI speakers and TVs.
Therefore, Xiao Ai Classmate can be directly applied to automobile products at present.
Mr. Povey, an internationally renowned scientist in the field of speech recognition and AI, and the father of Kaldi, an open source speech recognition software solution, joined in Xiaomi in October, 2019. At that time, Lei Jun promised to agree to any conditions he offered.
How important is Daniel Povey to Xiaomi?
Kaldi is the most popular open source speech recognition tool nowadays. Almost all speech recognition products developed by any company, whether in academia or private industry, depend on the underlying framework provided by Kaldi.
Under the leadership of Mr. Povey, the start-up team of “PyTorch-y” at Kaldi announced a new framework for deep learning, using weighted finite state automata in April 2020, which is mainly used for automatic speech recognition.
With Mr. Povey’s efforts, at least 20 new functions for the Xiao Ai Classmate were developed or upgraded last year.
In February this year, Xiaomi began to adjust its technical architecture by merging the originally independent AI department into the technical committee. Ye Hangjun, General Manager of the AI department and Vice Chairman of the AIoT Strategy Committee, instead became the chairman of the technical committee. Cui Baoqiu, former chairman of the technical committee, was transferred to become President of Xiaomi‘s Qinghe University.
Ye and Cui are both Mr. Povey’s bosses, and they were also in charge of Xiao Ai Classmate. While Xiaomi‘s autopilot team is in its preparatory stage now, the internal autopilot business will be reported to Ye. Therefore, Xiaomi‘s car-making business has already involved many managers of the AI department. Moreover, similar to Apple, Xiaomi attaches great importance to the status of AI when promoting automobile projects.
Therefore, Xiao Ai Classmate has a high probability of being applied to Xiaomi‘s vehicles.
Let’s imagine this picture. In the future, the owners of Xiaomi Auto can easily control all Xiaomi series products through Xiao Ai Classmate. They can remotely switch on or lock the vehicle, air conditioner, navigation, and finish automatic parking through any device equipped with Xiao Ai Classmate.
Autopilot Talent Team: DeepMotion × Microsoft Research Asia
On August 25th, Xiaomi officially announced that it had signed a share purchase agreement with autonomous driving company DeepMotion. Lin Bin, Co-Founder and Vice Chairman of Xiaomi, then posted a picture on Weibo and said: “I am very happy that several former Microsoft colleagues have joined Xiaomi Auto!”
This picture exposes Xiaomi‘s ambitions for autonomous driving and desire for DeepMotion’s unique talent.
Many people in the industry said that apart from Daniel Povey, DeepMotion is another treasure Xiaomi has found in the AI field. Xiaomi spent $77.37 million (500 million yuan) on the acquisition of DeepMotion, the biggest acquisition after the firm officially announced it would build cars. However, the expected valuation of DeepMotion was $200 million when the company sought financing in early 2020. Other autonomous driving firms such as Pony.ai, WeRide and Momenta bagged many rounds of financing worth of $1 billion. Acquiring DeepMotion is a real bargain for Xiaomi.
Another reason for regarding it as a real bargain is that the overall condition of DeepMotion is strong and balanced.
Cai Rui, DeepMotion’s CEO, Li Zhiwei, CTO, Yang Kuiyuan, chief scientist, and R&D Director Zhang Chi are all from Microsoft Research Asia, which tends to bag top talent in the field of AI algorithms in China.
Four of them have been focusing on deep learning and stereo vision at Microsoft Research Asia. Based on computer vision positioning, recognition and 3D scene reconstruction technology, they have a deep accumulation in AI technologies such as image recognition.
Moreover, as early as 2017, when Cai, Li and Yang were still at Microsoft Research Asia, they realized that the autonomous driving technology based on AI application would develop rapidly, and established their entrepreneurial direction.
Not long after the establishment of the company, DeepMotion built a scheme to help unmanned vehicles realize 3D perception, HD-Map drawing and positioning through camera array. A year later, DeepMotion achieved a mature ability to make HD maps of high-speed expressways on a large scale. The automatic drawing degree of HD maps for complex central city roads can reach 90%.
Therefore, DeepMotion can not only help Xiaomi to initially fill in the lineup of directors of each module of autonomous driving unit, but also lay the foundation for the R&D of autonomous driving technology used by Xiaomi Auto.
After the acquisition, there is a high probability that all members of DeepMotion will join Xiaomi. The backbone personnel will assume the role in charge of each module of automatic driving and report to Ye Hangjun.
Once the DeepMotion team is successfully integrated with the Xiaomi Auto team, it can basically be concluded that Xiaomi Auto’s automatic driving team will focus on vision, as well as high-precision map crowdsourcing, which is also the main direction of many well-known automatic driving companies at present.
For example, this applies to Horizon, Kuandeng Technology and Xpeng Motors. After Wang Tao, Co-Founder of Drive.ai, an autonomous driving company, became the head of Xpeng Motor’s perception R&D in North America, Xpeng also adopted the scheme mentioned above.
Autopilot Supply Chain: Hesai Technology × Geometrical Perception and Learning
On June 8, Hesai Technology, a Chinese lidar start-up, announced that it had completed financing worth over $300 million, and one of the leading investors was Xiaomi.
According to public data, Hesai Technology is a senior unicorn of lidar technology in China, and has deep accumulation in the underlying technology of lidar, core components, photoelectric chips, design process, automatic production, functional safety, active anti-jamming technology and perception ability based on deep learning.
The fresh funds will be used to support the mass production and delivery of hybrid solid-state lidar, the construction of Hesai Maxwell Intelligent Manufacturing Center, and the R&D of vehicle-level high-performance lidar chips.
More than a month later, Xiaomi took the lead in financing of another smart car supply chain company whose product is quite different from Hesai Technology – Geometrical Perception and Learning Co. Ltd., with an investment of nearly 400 million yuan.
Xue Dan, Founder and CEO of the company, is a member of the National Shanghai Artificial Intelligence Expert Committee, and once served as the chief engineer of an aerospace institute.
The core product of this start-up, which is positioned as tier one, is the integrated system of software and hardware for automatic driving, which adopts the technology integrating 4D millimeter wave imaging radar, visible light vision and infrared imaging.
The vehicle-level 4D millimeter wave radar adopted by G-PAL can achieve more than 6000 point clouds, a detection distance of more than 250 meters and a distance accuracy of 0.05 meters in a single frame. It can identify vehicles, non-motor vehicles, pedestrians, speed bumps, roadsides, guardrails and other objects, and provide outline information and behavior patterns of targets.
The goal of G-PAL is very clear. Taking 4D millimeter wave radar as the breakthrough point, G-PAL is trying to become an automatic driving solutions provider. Through independent R&D and manufacturing of software and hardware, it can provide car enterprises with automatic driving-oriented systems based on machine perception and deep learning.
Compared with Hesai Technology, G-PAL has chosen an easier route when trying to get involved in the field of vehicles. Perhaps this is exactly what Xiaomi values.
Injecting money into two smart car supply chain start-ups with completely different goals, reports show that Xiaomi still holds uncertainties about the future of its autonomous driving supply chain. But in any case, the supply chain is also a vital business component that must be firmly grasped.
Intelligent Transportation Ecological Chain: Zongmu Technology × AIpark
In addition to focusing on autonomous driving, in June and August this year, Xiaomi invested in two highly targeted “niche” technology companies: Zongmu Technology and AIpark.
On June 3, Zongmu Technology secured financing of $190 million, which was led by Xiaomi Changjiang Industrial Fund.
The fund’s founder, Tang Rui, was a senior to Wang Xing, the founder of Meituan, while studying together at Tsinghua University. He used to be the global senior engineering director of CSR (now acquired by Qualcomm). Around 2010, the CSR car navigation multimedia platform developed by him became the benchmark of China’s car navigation entertainment system market, with annual shipments reaching 10 million units.
In August 2012, Tang left CSR and founded Zongmu Technology, the leading ADAS technology supplier in China.
In October 2014, Zongmu Technology released the most comprehensive ADAS system solution in the industry at that time. In November 2017, the autonomous parking product of low-speed L4 autonomous driving was released. A year later, Zongmu cooperated with the Visteon Asia Pacific Company to jointly develop its own automatic parking technology.
At the end of 2018, Zongmu received a project notice from Hongqi, a sub-brand of FAW. Under this arrangement, would deploy L4-level autonomous parking services for low-speed automatic driving on the mass production models of Hongqi in 2020. Zongmu is also the first self-driving enterprise in China to take the lead in obtaining fixed contracts for L4-level mass production projects from automakers.
The direction of Zongmu’s exploration is to open up the whole autonomous parking ecology. It not only aims at liberating the manpower of car owners, but also bringing value-added services to shopping malls and automobile service industries.
At the end of August, Xiaomi strategically invested in the intelligent intercommunication company AIpark, which is the sixth company Xiaomi invested in since announcing its decision to begin car-making.
Yan Jun, the founder of AIpark, was also the founder of Wisdom Star Information Technology Co., Ltd., targeting the field of text big data. Through AI technology, it provides solutions for large-scale city-level intelligent parking.
The whole business model of AIpark is to connect urban traffic managers, parking service operators with car owners and users. Through apps facing consumers, it can directly inform car owners and users of current road conditions, the location of parking lots and parking spaces.
Under Xiaomi‘s investment, Zongmu Technology, which focuses on autonomous parking ecology, and AIpark, which focuses on city-level smart parking business, have fully being involved in the ecological chain of smart vehicles.
The speech of Jiang Wen, managing director of Xiaomi‘s strategic investment department, clearly stated how wide the firm’s car-making industry layout is. “Smart transportation and smart city have always been the investment fields that Xiaomi pays close attention to. AIpark has provided diversified product solutions for roadside public berths and commercial parking lots in nearly 30 cities, and has become the leader of the smart parking industry. We expect AIPark’s smart parking business to make steady progress, and then to closely connect with Xiaomi‘s smart ecology.”
New Energy Supply Chain: Svolt Energy × Ganfeng Lithium Battery
The third key line for Xiaomi to build cars is in the field of power batteries.
On July 30 of this year, Ganfeng Lithium announced that its subsidiary Jiangxi Ganfeng Battery Technology Co., Ltd., received investors’ capital increase, which was led by Xiaomi Changjiang Industrial Fund, with 187.5 million yuan. After the capital increase is completed, the shareholding ratio of Xiaomi Changjiang Industrial Fund is 3.51%.
Ganfeng Lithium Battery was established in 2013, and its parent company Ganfeng Lithium is the leader of the domestic lithium industry and the international lithium compound (lithium salt) deep processing industry. At present, its lithium hydroxide production capacity is 85,000 tons, ranking first in China.
Ganfeng has many lithium resources, such as lithium mines and salt lakes overseas, with a total amount of 27.61 million tons of LCE, ranking first in the world.
At the same time, as an important global supplier of lithium hydroxide, Ganfeng also cooperates with leading car companies such as Tesla, BMW and Volkswagen, as well as battery giants such as LG Chem.
Ganfeng has formed five lithium battery product systems, including consumer batteries, small polymer batteries, power batteries, energy storage batteries and solid-state batteries. The firm’s products are widely used in electric buses, electric ships, rail transit, electric forklifts, 5G intelligence, bluetooth headsets, medical devices and more.
Another big move by Xiaomi in the automotive energy field is to become an important shareholder in new energy firm Svolt Energy, with a capital contribution of 100 million yuan in its B round of financing.
Svolt Energy is currently in the first batch of enterprises in the world to overcome the technical difficulties of cobalt-free batteries, also realizing their vehicle application. ORA Cherry Cat, which was officially unveiled at the Chengdu Auto Show, is the world’s first mass-produced new energy vehicle equipped with cobalt-free batteries.
Cobalt is one of the most important scarce energy sources used to produce automobile batteries, while the proven reserves of cobalt in the world are only 7.1 million tons. In a sense, cobalt scarcity has begun to threaten the supply chain of Chinese EV makers, especially those that rely heavily on imports of the resource. Data show that in 2017, China imported 100,700 tons of cobalt worth $339,196,700 – an average price of $34 million per ton.
Therefore, how to reduce or even eliminate cobalt is a challenge for global battery enterprises. Nowadays, the only enterprise that can truly make lithium batteries cobalt-free is Svolt Energy. Although Tesla put forward its own cobalt-free goal in 2020, so far it has only reduced the cobalt content in its lithium batteries.
Just one month before Xiaomi injected capital investment into Svolt, the firm held a ceremony for the mass production of cobalt-free battery in Jiangsu, becoming the first power battery enterprise in the world to break through the technical difficulty. According to the plan, mass production of cobalt-free batteries will be finished at its Ma’anshan base before the end of the year.
In fact, Xiaomi is now aiming at the future of new energy batteries. In a sense, Svolt Energy broke the patent monopoly of overseas batteries and opened up a new path for the domestic lithium battery industry and the new energy automobile industry.
Xiaomi grasps not only the promising power of battery suppliers, but also of the upstream raw material suppliers.
The Basic Framework of Vehicle Building Is about to Be Settled
Over the past six months, Xiaomi has prepared related products, talent, as well as industrial investment and acquisition, to form a great basic framework for its vehicle building plan.
Firstly, at the beginning of this year, Xiaomi aimed at the semiconductor industry and invested in eight chip companies, including Yunyinggu Technology, BYD Semiconductor and Canxin Semiconductor, through the Xiaomi Changjiang Industry Fund based in the province of Hubei.
Secondly, the AI department, which created a Xiao Ai Classmate, was merged into the technical committee. The automobile project attached great importance to AI.
Thirdly, Xiaomi invested in the intelligent transportation ecological chain and the automatic driving supply chain.
Fourth, Xiaomi engaged with Svolt Energy and Ganfeng Lithium to ensure its preparation for the integration of energy and intelligence.
Last but not least, it invested in Alpark, equipped with autonomous parking services, and acquired DeepMotion with all its key talent.
In conclusion, Xiaomi has made a comprehensive investment and acquisition plan.
In addition, through Xiaomi‘s series of actions, we can see that there are a wide range of key parts that together comprise the field of autonomous driving and new energy in China, such as automatic parking, production-level lidar, cobalt-free batteries, maps and positioning, and 4D millimeter wave radar.
More and more technology startups are moving closer to the tech giants that suit them, looking for the best place to camp, instead of standing on their own feet.
Industry giants like Waymo, RoboTaxi, Pony.ai, Drive.ai, who spent a lot of money hoping to cover all aspects of the driverless vehicles field, have now become rarer.
This kind of environment seems to be suitable for Xiaomi today.
Taking news as the core, Jiemian provides all kinds of high-quality information, making it the first media outlet for China’s middle class.