# Wang Xingxing

> Source: https://aiwiki.ai/wiki/wang_xingxing
> Updated: 2026-10-30
> Fact-checked: 2026-10-30
> Categories: Chinese AI, People, Physical AI, Robotics
> License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) - attribute to "AI Wiki (aiwiki.ai)"
> Cite as: AI Wiki. "Wang Xingxing." aiwiki.ai, 30 Oct 2026. https://aiwiki.ai/wiki/wang_xingxing
> From AI Wiki (https://aiwiki.ai), the free encyclopedia of artificial intelligence. Reuse freely with attribution.

**Wang Xingxing** (Chinese: 王兴兴; born 1990) is a Chinese roboticist and entrepreneur who founded [Unitree Robotics](https://aiwiki.ai/wiki/unitree) in Hangzhou in August 2016 and is its chairman, general manager and chief technology officer.[6] He began building legged robots as a graduate student at Shanghai University, where his low-cost, electric-motor-driven quadruped XDog drew enough attention in 2016 to persuade him to leave a new job at the drone maker [DJI](https://aiwiki.ai/wiki/dji) after about two months and start the company.[7][9][10] Unitree went on to become a large-volume seller of [quadruped robots](https://aiwiki.ai/wiki/quadruped_robot) and [humanoid robots](https://aiwiki.ai/wiki/humanoid_robot), and Wang remained its controlling shareholder when the company listed on the Shanghai Stock Exchange's STAR Market in August 2026.[6][19]

Wang became a national public figure in China in early 2025, after Unitree's H1 humanoids danced at the CCTV Spring Festival Gala and he was one of six entrepreneurs who spoke at Xi Jinping's symposium on private enterprises in Beijing on February 17, 2025.[11][12] In October 2026, at a Unitree-hosted conference in Hangzhou, he described his team's "zero training" experiments controlling robots with what 36Kr identified as OpenAI's [GPT-6 Astra](https://aiwiki.ai/wiki/gpt_6_astra) (other reports said only "GPT-6") and predicted that [physical AI](https://aiwiki.ai/wiki/physical_ai) could see a major breakthrough within one to three years.[1][2]

## At a glance

| Item | Detail |
|---|---|
| Native name | 王兴兴 |
| Born | 1990, Zhejiang province (Ningbo; Yuyao according to The Standard)[6][7][8] |
| Education | Undergraduate study in mechatronics engineering, Zhejiang Sci-Tech University (enrolled 2009); master's degree, Shanghai University (enrolled 2013, graduated 2016)[6][7] |
| Before Unitree | About two months at DJI in Shenzhen, 2016[7][9] |
| Current roles | Chairman, general manager and chief technology officer of Unitree (since August 2016); core technical staff member[6] |
| Ownership of Unitree | 21.4395% directly after the August 2026 offering; 65.3090% of voting rights through dual-class shares and an employee platform he controls[6] |
| Other positions | Member of the Hong Kong Chief Executive's Council of Advisers (term July 1, 2025 to June 30, 2027)[15] |
| Recognition | Fortune China 40 under 40 (2023); TIME100 AI (2025)[14][15] |

## Early life and education

Wang was born in 1990 in Ningbo, Zhejiang. The South China Morning Post, drawing on Chinese media reports, wrote that he built model aircraft as a child and was assembling miniature turbojet engines by middle school, but that he struggled with English, a compulsory subject in China's exam-driven system.[7] The Hong Kong newspaper The Standard gives his hometown as Yuyao, a county-level city in Ningbo, and describes his family as modest.[8] Unitree's prospectus confirms the 1990 birth year and lists his education as a master's degree.[6]

In 2009 he enrolled at Zhejiang Sci-Tech University in Hangzhou to study mechatronics engineering. As a first-year student he built a small bipedal robot that could walk a few steps on a budget of about 200 yuan. He received his first patent, for a "multi-force feedback device for fingers," in 2011, and his graduation thesis concerned a brushless DC motor controller.[7] Wang later told TIME that his preference for simple designs with few parts was formed when he built that first robot on a shoestring budget.[14]

He began a master's degree at Shanghai University in 2013. According to the SCMP, a poor English exam score again kept him out of a top-tier university.[7] Wang told KrASIA in 2021 that he started building quadrupeds as a postgraduate because most agile four-legged robots of the time were hydraulic, large and expensive, and he saw that battery-powered electric motors could make them cheaper and more agile.[9]

## XDog and DJI

XDog was the quadruped Wang built as a graduate student. Unitree describes it as a full-degree-of-freedom quadruped driven by low-cost external-rotor brushless motors, for which Wang handled the control simulation, motor selection, motor-driver boards and driver software, master-slave control architecture, leg design and control programs.[10] The SCMP reported that he spent about 20,000 yuan on the prototype in 2015 and won 80,000 yuan in prize money with it in a local competition.[7] KrASIA reported that XDog, developed for his master's thesis, could walk, run and jump, became an internet sensation, and attracted buyers and investors.[9]

After graduating in 2016 Wang joined DJI in Shenzhen. In his account to KrASIA: "I started the company after working for two months at DJI. I didn't have enough financial resources or strong connections. For the first three years, there were times when we could not even pay the salaries of our employees. Thankfully, we had revenues at the end of 2018 and managed to get more funding from investors."[9] Unitree's own history says that media coverage of XDog in June 2016 led Wang, then still working in Shenzhen, to resign and found the company with angel investment.[10] The SCMP put that first angel funding at 2 million yuan.[7]

## Founding and running Unitree

Unitree's predecessor, Hangzhou Yushu Technology Co., Ltd., was registered on August 26, 2016, with registered capital of 100,000 yuan contributed entirely by Wang. He has been chairman, general manager and chief technology officer since that month.[6] (KrASIA's 2021 article dated the launch to May 2016; the registration date comes from the company's prospectus.)[6][9] According to the SCMP, the name combines "universe" and "tree" and alludes to the gaming phrase "lighting up the technology tree."[7]

The company rebuilt XDog as the Laikago quadruped, which appeared in September 2017.[10] Later quadrupeds included the A1 (2020) and Go1 (2021), and Unitree moved into humanoids with the [H1](https://aiwiki.ai/wiki/unitree_h1) in August 2023, followed by the [G1](https://aiwiki.ai/wiki/unitree_g1) in May 2024, the [R1](https://aiwiki.ai/wiki/unitree_r1) in July 2025 and the [H2](https://aiwiki.ai/wiki/unitree_h2) in October 2025.[10] In 2021 Wang described Unitree's strategy to KrASIA as consumer-oriented and fast-cycling, saying of [Boston Dynamics](https://aiwiki.ai/wiki/boston_dynamics): "Boston Dynamics' robot dogs are undeniably more capable than ours in terms of functionality, but it is not our direct competitor. It took them five years to make Spot available for purchase, while we have released at least one or two products every year since our launch."[9]

Unitree's marketing director told the SCMP in early 2025 that Wang was deeply involved in research and development, personally interviewed job candidates and oversaw nearly every aspect of operations.[7] In an August 2025 People's Daily interview Wang said, "I oversee tech and products."[13]

## National profile, 2025

On January 28, 2025, Unitree H1 humanoids performed a yangge folk-dance routine titled "Yang BOT" alongside human dancers at the CCTV Spring Festival Gala, in a program directed by Zhang Yimou with the Xinjiang Arts Institute.[12][10] Wang told the People's Daily that the scale of the reaction was "completely unexpected," since Unitree's "robot cow" quadrupeds had appeared on the 2021 gala and the company had expected a similar impact.[13]

On February 17, 2025, Wang was among six private entrepreneurs, with Ren Zhengfei of Huawei, Wang Chuanfu of BYD, Liu Yonghao of New Hope, Yu Renrong of Will Semiconductor and [Lei Jun](https://aiwiki.ai/wiki/lei_jun) of Xiaomi, who offered opinions and advice at a symposium on private enterprises attended by Xi Jinping in Beijing.[11] The SCMP noted that the 35-year-old was seated in the front row and was the youngest of a group of prominent attending executives that included Jack Ma and Pony Ma.[7] Afterward he told CCTV: "Artificial-intelligence-driven robots are evolving at an incredibly fast pace, surpassing my expectations. Every day brings new surprises. I believe that by the end of the year, humanoid robots will reach a whole new level."[7]

Later that year the Hong Kong government announced on June 27, 2025, that it had appointed Wang as a new member of the Chief Executive's Council of Advisers, for a term from July 1, 2025, to June 30, 2027. The government's biography identified him as Unitree's founder, chief executive officer and chief technology officer and noted that he had been named one of Fortune China's 40 business leaders under 40 in 2023.[15] In 2025 TIME included him on its third annual TIME100 AI list.[14]

## IPO and ownership

Wang led Unitree through its STAR Market initial public offering, which is covered in detail in [Unitree Robotics IPO](https://aiwiki.ai/wiki/unitree_ipo_2026). The final prospectus shows that before the offering he directly held 86,714,964 shares, or 23.8216 percent of the company, and controlled 68.7816 percent of the voting rights. That figure came from a dual-class arrangement adopted on May 16, 2025, which made 44,074,296 of his shares Class A shares carrying ten votes each, together with his control of the Shanghai Yuyi employee equity incentive platform. After the offering his direct stake fell to 21.4395 percent and his controlled voting rights to 65.3090 percent.[6] The prospectus also lists his 2025 compensation from Unitree as 2.3858 million yuan.[6]

The shares began trading in Shanghai on August 19, 2026, and rose 460 percent on their first day.[19] The Standard reported that, counting direct and indirect holdings of about 30.02 percent, Wang's stake was briefly worth more than 133.5 billion yuan during that first session and more than 80 billion yuan at the close, which the paper said made him the richest entrepreneur born after 1990, measured against New Fortune magazine's 2025 rankings.[8] These are paper values tied to a single day's share price.

The next day Wang gave a keynote speech titled "From Exhibits to Products: The Next Decade of the Humanoid Robot Industry" at the main forum of the 2026 [World Robot Conference](https://aiwiki.ai/wiki/world_robot_conference). He said the industry's biggest bottleneck was the weak generalization of embodied intelligence, and he publicly described for the first time an internal "physical AI robot self-evolution" project.[17]

## Views on AI and robotics

Wang has consistently described AI capability, not hardware, as the limit on robot adoption. The timelines below refer to different milestones and are not directly comparable.

| Date | Setting | What Wang said |
|---|---|---|
| February 2025 | CCTV interview after the Xi symposium | Robots were evolving faster than he expected; humanoids would "reach a whole new level" by the end of 2025[7] |
| August 2025 | People's Daily interview | The main hurdle is that robot AI "isn't advanced enough to break the critical threshold"; "notable progress might come in three to five years," and entering homes would take longer because of ethical and safety concerns[13] |
| 2025 | TIME100 AI profile | "To truly put AI to work and help solve real-world problems for humanity, we need robotics"; "AI and robotics are inseparable"[14] |
| August 20, 2026 | World Robot Conference keynote | An embodied-AI "ChatGPT moment" would arrive when a robot taken into an unfamiliar home can do about 80% of tasks; "if fast, perhaps two to three years; if slow, five or ten years"[17] |
| October 22, 2026 | World Physical AI Conference, Hangzhou | Physical AI may see a major breakthrough "in one to two years if fast, two to three years if slow"[1][2] |

In the August 2026 keynote Wang attributed the generalization problem to imperfect alignment between AI models and the physical world. A language model's inputs and outputs stay inside a vector space without loss, he said, while every physical input and output on a robot adds deviation and information loss, so that broad instructions are executed well but the "last few centimeters, or last few millimeters" go wrong. He called this "a problem that will certainly be solved in the next few years."[17]

## October 2026 World Physical AI Conference speech

On October 22, 2026, Unitree held the "2026 World Physical AI Conference and Unitree Partner Ecosystem Conference" (2026世界物理AI大会暨宇树合作伙伴生态大会) in Hangzhou, the first partner ecosystem conference in the company's ten-year history. Wang gave a speech titled 《相信，尽可能相信AI》 ("Believe: believe in AI as much as you can"; AI Wiki translation).[1] The tech outlet Zhidongxi reported that the speech lasted about 30 minutes.[2] Other guests included [Wang Jian](https://aiwiki.ai/wiki/wang_jian), founder of Alibaba Cloud and director of Zhejiang Lab, and Yao Ming, who held a roundtable conversation with Wang.[1][2]

Wang opened by saying: "全世界几乎所有人，对AI的重视程度和信仰还不够，包括我自己" ("Almost everyone in the world, myself included, still does not take AI seriously enough or believe in it enough"; AI Wiki translation).[1] He said he had recommended AI-related stocks to colleagues four or five years earlier and that some Unitree employees' personal wealth could have grown tenfold or more had they believed more in AI. He argued that ordinary people have not yet felt AI's effects because a technology typically takes five to ten years to spread after its breakthrough.[1]

### GPT-6 Astra experiments

According to 36Kr, Wang said that the recently released GPT-6 Astra had given the industry a huge shock ("给了行业巨大的冲击") and that Unitree had tried the model in several robot experiments.[1] As 36Kr described them:

- **Painting.** Unitree did not write complex programs for the task and used only prompts (提示词) to have a robot paint. The robot also changed paints and cleaned its brush on its own.[1]
- **Go.** The robot recognized the Go board and placed stones accurately in the intended positions.[1]

36Kr reported that Wang "repeatedly emphasized" that these experiments were done under "零训练" ("zero training") conditions.[1] Other reports of the same speech referred to the model only as "GPT-6." IT Home quoted Wang as saying the placement accuracy was "very high," and Zhidongxi's account added Chinese chess and autonomously finding a kitchen and locating tools to the list of tasks.[2][3] These are Wang's own descriptions of internal tests.

Wang also described limits. In 36Kr's summary, he said large models already handle robot task planning and navigation fairly well but still make mistakes in actual manipulation: picking up an object is relatively easy, while assembling precision parts is much harder because the robot must adjust its movements in real time to an object's position and the forces involved. In his view current models still fall clearly short at perceiving changes in the real environment and correcting errors promptly.[1] He identified aligning AI models' inputs and outputs with real robots as the core challenge before embodied AI reaches its "ChatGPT moment," and Zhidongxi reported him as saying that planning had been solved.[1][2]

### Predictions and plans

Wang said physical AI may see a major breakthrough "快的话一到两年，慢的话两到三年" ("in one to two years if fast, two to three years if slow"; AI Wiki translation), and that large models may eventually "击穿" ("break through," literally "pierce") physical AI.[1] 36Kr's headline condensed this to "two or three years."[1] Zhidongxi quoted him as saying: "We are heading toward the ChatGPT moment of embodied intelligence: in 80 percent of unfamiliar settings, robots can successfully complete about 80 percent of tasks through voice or text instructions" (AI Wiki translation of the Chinese quotation).[2] He said he hoped that by the end of 2026 or in 2027 a person could tell a robot to tidy a room and it would do so.[1][2] IT Home quoted a slightly more cautious version: robots handling about 80 percent of tasks through voice and language instructions "if fast, there is a chance next year; to be safe, perhaps the year after."[3]

He also outlined Unitree's plans:

| Topic | What Wang said |
|---|---|
| Self-evolution model | Unitree is developing "物理AI机器人自进化V1" (physical AI robot self-evolution V1), in which AI agents write code, train and test in simulation, verify on real robots and feed results to an automated evaluation system in a closed loop. 36Kr noted that training efficiency, skill transfer and real-world success rates remained unverified.[1] Kuai Technology and IT Home, describing the same self-evolution system, added that the AI models would also gather research papers themselves.[3][4] |
| Models | Since 2025 Unitree has used pretrained deep [reinforcement learning](https://aiwiki.ai/wiki/reinforcement_learning) to build a pretrained whole-body motion control model; it also works with Chinese and foreign large-model companies.[1] |
| Production | The G1 has had about 10,000 to 20,000 units produced since its May 2024 launch, which Wang called the world's best-selling humanoid; he said the industry will eventually need capacity to build about ten million robots a year.[1][2] |
| Industry priorities | Large-scale robot manufacturing, development and deployment of physical AI models, very large compute servers, and compute and memory chip manufacturing.[1] |
| Scarce resources | Data, compute, chip production capacity, energy and minerals.[1] |
| Ecosystem | A "宇创计划" (Yuchuang Plan) startup program offering capital, robot platforms, tools, engineering and application scenarios, with a goal of incubating more than 200 ecosystem companies in five years.[1] |

In the roundtable with Yao Ming, Wang defended Unitree's dancing, fighting and running demonstrations as training for practical work: "如果你连自己走路都走不好，跑都跑不起来，那你还指望这样的机器人去干活吗？" ("If a robot can't even walk properly or run, how can you expect it to do any work?"; AI Wiki translation). He added that fighting tests reliability through collisions and falls and demands complex motion control.[1]

He ended with personal remarks. He said this generation has a real chance to achieve [superintelligence](https://aiwiki.ai/wiki/superintelligence) or [AGI](https://aiwiki.ai/wiki/artificial_general_intelligence) and to raise productivity through humanoid robots, that competition in the industry is "非常惨烈" ("extremely brutal"), and that "现在是最好的时代，我非常害怕浪费每一天" ("This is the best of times. I am very afraid of wasting a single day"; AI Wiki translation).[1][4]

### English-language circulation

On October 24, 2026 (UTC), Kyle Chan, whose X profile identifies him as a fellow at the Brookings Institution, posted an English rendering of the GPT-6 Astra passage of the 36Kr article under the comment "Unitree's Wang Xingxing got AGI-pilled by Astra."[5] His translation closely follows the Chinese text but does not include Wang's stated timeframe, rendering the "击穿" line as large models being able to "break through the fundamental barriers holding back physical AI."[1][5]

## Research

Wang is listed as a co-author, under the name Xingxing Wang with a Unitree Robotics affiliation, of "OmniXtreme: Breaking the Generality Barrier in High-Dynamic Humanoid Control" (arXiv, February 2026). The paper comes from a joint laboratory of [BIGAI](https://aiwiki.ai/wiki/bigai) and Unitree. It trains a flow-matching policy on a large motion library, refines it with actuation-aware reinforcement learning, and reports flips and acrobatics on a Unitree G1.[18]

## Recognition

- Fortune China 40 business leaders under 40, 2023, as cited by the Hong Kong government.[15]
- TIME100 AI, 2025.[14]
- TIME magazine cover, July 2026. According to China Daily, TIME described the cover interview, which featured Unitree's rideable GD01 mech, as Wang's first interview with an international media outlet.[16]

## See also

- [Unitree Robotics](https://aiwiki.ai/wiki/unitree)
- [Unitree Robotics IPO](https://aiwiki.ai/wiki/unitree_ipo_2026)
- [GPT-6 Astra](https://aiwiki.ai/wiki/gpt_6_astra)
- [Physical AI](https://aiwiki.ai/wiki/physical_ai)
- [Embodied AI](https://aiwiki.ai/wiki/embodied_ai)
- [Humanoid robot](https://aiwiki.ai/wiki/humanoid_robot)

## References

1. Qiu Xiaofen (邱晓芬). ["宇树王兴兴：物理AI会在未来两三年突破，非常害怕浪费每一天"](https://www.36kr.com/p/4037854039610505) (Unitree's Wang Xingxing: physical AI will break through in the next two or three years; very afraid of wasting each day). 36Kr, October 2026 (page timestamp October 23, 2026). Accessed October 30, 2026.
2. Zhidongxi (智东西). ["宇树王兴兴：全世界对AI的信仰都还不够，包括我自己"](https://t.cj.sina.com.cn/articles/view/5213469505/136bf3b41020016jpy) (Unitree's Wang Xingxing: the world's belief in AI is still not enough, including my own). Report dated October 22, 2026; Sina Finance copy posted October 26, 2026. Also republished by PingWest: https://www.pingwest.com/a/318659
3. IT Home (IT之家). ["宇树科技王兴兴：我很害怕浪费每一天"](https://tech.ifeng.com/c/8xQnrfic2hF) (Unitree's Wang Xingxing: I am very afraid of wasting each day). Via ifeng, October 22, 2026.
4. Kuai Technology (快科技). ["宇树科技王兴兴：真的非常害怕浪费每一天"](https://tech.ifeng.com/c/8xQjXMYhJUq) (Unitree's Wang Xingxing: truly very afraid of wasting each day). Via ifeng, October 22, 2026.
5. Kyle Chan (@kyleichan). [Post on X](https://x.com/kyleichan/status/2113798832286753050). October 24, 2026 (UTC).
6. Yushu Technology Co., Ltd. ["宇树科技股份有限公司首次公开发行股票并在科创板上市招股说明书"](https://static.sse.com.cn/disclosure/listedinfo/announcement/c/new/2026-08-14/688836_20260814_AJKD.pdf) (Prospectus for the initial public offering and STAR Market listing). Shanghai Stock Exchange, August 14, 2026.
7. Wency Chen. ["Meet Wang Xingxing, the young Chinese robotics star from Unitree at Xi Jinping's symposium"](https://www.yahoo.com/news/meet-wang-xingxing-young-chinese-093000107.html). South China Morning Post (via Yahoo News), February 21, 2025.
8. ["Unitree Robotics founder Wang Xingxing tops post-90s rich list with 133.5b yuan stake"](https://www.thestandard.com.hk/finance/article/340376/Unitree-Robotics-founder-Wang-Xingxing-tops-post-90s-rich-list-with-1335b-yuan-stake). The Standard (Hong Kong), August 19, 2026.
9. Stephanie Li. ["Unitree Robotics develops personal robot dogs that can jog with you"](https://letschuhai.com/unitree-robotics-develops-personal-robot-dogs-that-can-jog-with-you). KrASIA, July 2021 (36Kr mirror).
10. Unitree Robotics. ["About Unitree"](https://www.unitree.com/about/). Company history page. Accessed October 30, 2026.
11. Xinhua. ["Xi urges healthy, high-quality development of private sector"](https://english.www.gov.cn/news/202502/17/content_WS67b3240fc6d0868f4e8efb3e.html). Via english.gov.cn, February 17, 2025.
12. ["Unitree's humanoid robots steal the show at 2025 CCTV Spring Festival Gala with AI-powered Yangge dance"](https://technode.com/2025/01/29/unitree-robotics-humanoid-robots-steal-the-show-at-2025-cctv-spring-festival-gala-with-ai-powered-yangge-dance/). TechNode, January 29, 2025.
13. Dou Hao. ["Unitree founder Wang Xingxing to People's Daily"](https://www.ehangzhou.gov.cn/2025-08/14/c_294924.htm). People's Daily (via eHangzhou), August 14, 2025.
14. ["Hangzhou tech company CEOs make TIME100 AI list"](https://www.ehangzhou.gov.cn/2025-09/02/c_294907.htm). China Daily (via eHangzhou), September 2, 2025, reproducing TIME's profile text.
15. Government of the Hong Kong SAR. ["Appointments to second term of Chief Executive's Council of Advisers announced"](https://www.info.gov.hk/gia/general/202506/27/P2025062700529.htm), June 27, 2025; biography annex: https://gia.info.gov.hk/general/202506/27/P2025062700523_500359_1_1751008941023.pdf
16. ["China Daily: Unitree Robotics founder featured in TIME magazine"](https://english.sse.com.cn/news/newsrelease/voice/c/c_20260729_10827108.shtml). Shanghai Stock Exchange (reprinting China Daily), July 29, 2026.
17. Liu Chenyin (刘晨茵). ["王兴兴透露宇树新方向：物理AI机器人自进化 让机器人自己为自己升级"](https://tidenews.com.cn/news.html?id=3535379) (Wang Xingxing reveals Unitree's new direction: physical AI robot self-evolution). Tide News (潮新闻), August 20, 2026.
18. Yunshen Wang, Shaohang Zhu, Peiyuan Zhi, Yuhan Li, Jiaxin Li, Yong-Lu Li, Yuchen Xiao, Xingxing Wang, Baoxiong Jia and Siyuan Huang. ["OmniXtreme: Breaking the Generality Barrier in High-Dynamic Humanoid Control"](https://arxiv.org/abs/2602.23843). arXiv:2602.23843, February 27, 2026.
19. Nicholas Gordon. ["Unitree, famous for its dancing robots, surges by 460% on its trading debut, lifting valuation to $66 billion, far ahead of U.S. competitors"](https://fortune.com/2026/08/19/unitree-china-dancing-robots-ipo-trading-surge-valuation/). Fortune, August 19, 2026.