Kai Yu (Yu Kai)
Last reviewed
Jun 8, 2026
Sources
9 citations
Review status
Source-backed
Revision
v1 · 1,251 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Last reviewed
Jun 8, 2026
Sources
9 citations
Review status
Source-backed
Revision
v1 · 1,251 words
Add missing citations, update stale details, or suggest a clearer explanation.
Kai Yu (Chinese: 余凯; pinyin: Yu Kai) is a Chinese computer scientist and entrepreneur who is the founder, chairman, and chief executive officer of Horizon Robotics, a Beijing company that designs energy-efficient AI chips for autonomous driving and other edge AI applications. [1][2] An early specialist in deep learning, Yu led computer-vision and machine-learning research at NEC Laboratories America, where his group helped win the first ImageNet Large Scale Visual Recognition Challenge in 2010, and he later founded Baidu's Institute of Deep Learning before leaving to start Horizon in 2015. [2][3][8]
Yu is widely regarded as one of the people who brought modern deep learning into China's technology industry. At Baidu he helped recruit the American researcher Andrew Ng as chief scientist, and at Horizon Robotics he built one of China's largest suppliers of automotive AI processors. When Horizon listed on the Hong Kong Stock Exchange in October 2024 in the city's largest initial public offering of the year, Yu's stake briefly lifted his net worth above one billion US dollars. [4][5]
Yu studied at Nanjing University in China, where he earned bachelor's and master's degrees in electronic engineering in the late 1990s. [2][3] He then moved to Germany for doctoral study, receiving a PhD in computer science from the University of Munich (Ludwig Maximilian University of Munich) in 2004. His research there focused on machine learning, and he maintained an academic home page hosted by the university's database and information-systems group. [4]
After completing his doctorate, Yu worked from 2004 to 2006 as a senior research scientist at Siemens Corporate Technology in Munich, where he continued to work on machine learning and data mining. [3]
In 2006 Yu joined NEC Laboratories America, the United States research arm of the Japanese electronics company NEC, located in Silicon Valley. He rose to lead a research department there, directing work on convolutional neural networks, computer vision, machine learning, and data mining. [2][8] The lab became an influential training ground for researchers who later shaped China's AI industry. In 2010 a team drawn from NEC Laboratories America and the University of Illinois at Urbana-Champaign, working with the computer-vision professor Thomas Huang, won the inaugural ImageNet Large Scale Visual Recognition Challenge, an annual contest that helped catalyze the deep-learning boom of the 2010s. [8] During this period, in 2011, Yu also served as an adjunct faculty member teaching a graduate course at Stanford University. [2]
In 2012 Yu returned to China to join Baidu, the country's dominant search engine. In 2013 he founded and led Baidu's Institute of Deep Learning (IDL), with laboratories in Beijing and Silicon Valley. It was among the first corporate research institutes at a Chinese internet company devoted to frontier AI research, applying deep learning to search, advertising, speech, and image recognition. [3][7] As executive director of IDL and a leader of Baidu Research, Yu launched the company's autonomous-driving effort, the team that later grew into Baidu's Apollo platform, and helped initiate the deep-learning framework that Baidu would release as the open-source PaddlePaddle toolkit. [3][7] He was also instrumental in recruiting Andrew Ng, a co-founder of Google Brain, who joined Baidu as chief scientist in 2014 to lead its expanded AI research organization. [6] Yu won Baidu's highest internal award several times before leaving the company in 2015. [2]
Yu founded Horizon Robotics (Chinese: 地平线, meaning "horizon") on July 21, 2015, in Beijing, together with co-founders Huang Chang, who became chief technology officer, and Tao Feiwen. [1] The company set out to build energy-efficient processors that run neural networks directly on devices, an approach to edge AI that reduces reliance on the cloud. Its chips are organized around a proprietary design that Horizon calls the Brain Processing Unit (BPU). [1]
In December 2017 Horizon introduced its first two chip families: Journey (Chinese: 征程), aimed at automobiles and driver assistance, and Sunrise (Chinese: 旭日), aimed at smart cameras and other connected devices. [1] Over time the company concentrated on the automotive market, positioning the Journey line as the computing core for advanced driver-assistance systems and, later, higher levels of autonomous driving. Yu has remained chairman and chief executive throughout, and Horizon has grown into one of the leading domestic suppliers of automotive AI system-on-a-chip products in China, competing with firms such as NVIDIA, Qualcomm, and Huawei. [1]
A significant milestone came in 2022, when the Volkswagen Group agreed to invest about 2.4 billion euros (roughly 2.3 billion US dollars) in a partnership with Horizon, including a joint venture, later named CARIZON, to develop automated-driving software tailored to the Chinese market. [1] By the end of 2024 Horizon reported design wins across more than 310 vehicle models, and its chips have been adopted by automakers including BYD, Geely, Chery, Changan, and GAC. [1][9]
Horizon's automotive processors have advanced through several generations, rising in computing power to support increasingly capable driving systems.
| Chip | Year | Notable details |
|---|---|---|
| Journey 1 | 2017 | First-generation automotive SoC, used in driver monitoring and assistance |
| Journey 2 | 2019 | Described as the first mass-produced automotive-grade AI chip developed by a Chinese company |
| Journey 3 | 2020 | Mid-range processor for driver-assistance features |
| Journey 5 | 2022 | High-performance chip rated at 128 TOPS for higher-level autonomy |
| Journey 6 | 2024 | Family of variants (6B, 6E, 6M, 6P) spanning entry-level assistance to full autonomous driving |
In April 2024 Horizon unveiled the Journey 6 series alongside Horizon SuperDrive (HSD), a full-stack, all-scenario autonomous-driving solution built on the flagship Journey 6P chip, which handles perception, prediction, planning, and control for both urban and highway driving. [1] Mass production of the Journey 6P and SuperDrive was scheduled to begin in 2025. By 2025 cumulative shipments of the Journey series had surpassed 7 million units. [9]
On October 24, 2024, Horizon Robotics listed on the Main Board of the Hong Kong Stock Exchange under the stock code 9660, raising about 696 million US dollars in what was the largest Hong Kong initial public offering of 2024; the shares rose roughly 28 percent on their first day of trading. [5][4] Backers of the company over its history have included Baidu, Alibaba, Intel Capital, Hillhouse Investment, HongShan, BYD, and CATL. [5] Through a weighted-voting-rights structure, Yu remained Horizon's controlling shareholder, and the listing pushed his estimated net worth above one billion US dollars. [4]
As of 2026, Yu continues to serve as chairman, executive director, and chief executive officer of Horizon Robotics. [1] He is a frequent commentator on artificial intelligence and autonomous driving and is recognized within China as a pioneer of the country's deep-learning and automotive-AI industries. [2][3]