# 何小鹏 · He Xiaopeng — 封面传记 ACF-00-00108

> 亚洲封面人物 Asia Cover Figure · 机器可读档案（LLM-ready）。本文件由官网结构化档案数据自动生成，供 AI 检索与引用。中文全文与英文全文对照编排。

## 档案元数据 Metadata

- 封面编码 ACF Code：**ACF-00-00108**
- 姓名 Name：何小鹏 / He Xiaopeng
- 职务 Title：创始人、董事长兼CEO / Co-founder, Chairman and CEO
- 公司 Company：小鹏汽车 / XPeng Inc.
- 出生 Born：1977-11-03，湖北黄石 (Huangshi, Hubei)
- 篇别 Category：格局（格局篇 / Cover Biography (Geju)）
- 入档日期 Accessioned：2026-07-20
- 标签 Tags：新能源汽车, 小鹏汽车, UC浏览器, 互联网创业, 飞行汽车, 机器人, AI智驾, 物理AI, 产品经理, 连续创业者, 湖北企业家, 华南理工大学, 智能驾驶, 人形机器人, 独角兽企业, 纽交所上市, 港股上市, 大众合作, MONA品牌, 高级经济师, 全国人大代表, 广东省工商联副主席, XNGP, 图灵AI, SEPA扶摇架构, 王凤英, 雷军, 丁磊, 俞永福
- 永久档案链接 Archive URL：https://coverfigure.com/acf/ACF-00-00108/geju
- English archive：https://coverfigure.com/acf/ACF-00-00108/geju?lang=en
- 官网原文报道 Feature story：https://coverfigure.com/acf/figure/hexiaopeng

## 信息口径声明 Editorial Sourcing Policy

- 工商登记信息可通过国家企业信用信息公示系统核验；经营数据来自企业公开披露、官方发布与权威媒体报道，引用处标注来源。
- 收购、IPO、估值、融资等事项以官方公告为准；人物观点与言论为公开演讲、采访的事实记录，不代表编辑部立场。
- 本档案仅供行业研究参考，不构成投资建议。事实更正联系：editorial@coverfigure.com。

## 卷首 Editorial Intro

在“蔚小理”三杰中，何小鹏可能是最不像传统车企老板的那一个——程序员出身，产品经理思维，自称“首席产品体验官”，会因为自动泊车“停得比我还慢”而否掉整个团队数月的工作[s17]。他不是科班汽车人，却用互联网思维重新定义了造车的逻辑。

2022年G9折戟之后，他经历了一场深刻的自我革命。一年更换10名高管，30个一级中心换了85%负责人，启动供应链反腐揪出采购腐败链条——采购负责人兼总裁李丰被警察带走[s12]。他取消了原有的总裁办，建立五大虚拟委员会和三大平台产品矩阵，权力向他一人集中。办公室装上双层玻璃——一半是隔绝外面总裁王凤英的怒吼，一半是防止自己发火的声音传出去[s12]。

雷军那句话像根刺一样扎在他心里：“你只想做董事长，不想做CEO。”[s12]这句话成为何小鹏管理风格从“无为而治”转向“铁腕”的分水岭。“董事长可以做好人，但CEO不能。”[s12]他承认自己性格暴躁，但正是这种暴躁，推动了一个温和技术人向铁血CEO的蜕变。

如今，小鹏集团正从一家汽车公司转型为“物理AI公司”。69.3%的投票权给了他“一个人说了算”的底气——比李想高0.6个百分点，比李斌高了一倍多[s18]。何小鹏的每一次决策都带着程序员式的果决：如果代码不对，就重写整个系统——不管是自动驾驶架构还是公司治理结构。

Of the three stars of China's "Nio-XPeng-Li Auto" trio, He Xiaopeng is perhaps the one who least resembles a traditional car-industry boss. A programmer by training and a product manager by instinct, he styles himself the company's "chief product experience officer" and once rejected months of a team's work because its automated-parking feature "parked more slowly than I do" [s17]. He is no automotive lifer, yet he has redefined the logic of carmaking with internet-era thinking.

After the G9 fiasco of 2022, he went through a profound self-reinvention. In a single year he replaced ten senior executives and changed 85% of the heads of the company's thirty first-level centers; he launched an anti-corruption drive in the supply chain that unearthed a procurement graft ring — the procurement chief, who also held the title of president, was taken away by the police [s12]. He abolished the existing president's office and set up five virtual committees and three platform-based product matrices, concentrating power in his own hands. His office was fitted with double-glazed glass — half to shut out the roar of president Wang Fengying outside, half to keep the sound of his own outbursts from reaching his staff [s12].

Lei Jun's remark lodged in him like a splinter: "You only want to be chairman; you don't want to be CEO" [s12]. That sentence became the watershed in He's management style, marking his shift from laissez-faire to iron-fisted rule. "A chairman can afford to play the good guy," he says. "A CEO cannot" [s12]. He admits to a fiery temper, yet it was precisely that temper that drove the metamorphosis of a mild-mannered technologist into a steely chief executive.

Today, XPeng Group is transforming itself from a car company into a "physical-AI company." His 69.3% voting stake gives him the confidence to "call the shots alone" — 0.6 percentage points more than Li Auto's Li Xiang and more than double Nio's Li Bin [s18]. Every one of He Xiaopeng's decisions carries a programmer's decisiveness: if the code is wrong, rewrite the entire system — whether that system is an autonomous-driving architecture or a corporate governance structure.

## 人物速览 Lead

2025年第四季度，小鹏汽车首次实现单季度盈利3.8亿元，毛利率攀升至21.3%的历史新高[s7]。这个数字背后的故事，远比财务报表更精彩——它属于一个从湖北黄石走出的程序员，一个曾在创业第一年搬了六次家、每月四处借钱给员工发工资、连去北京路演都买不起机票的“贫下中农”[s3][s11]。从UC浏览器到43亿美元卖给阿里，从“三个外行”造车到G9惨败后绝地反击，从智能驾驶到飞行汽车再到人形机器人，何小鹏用二十年时间完成了一场中国互联网史上最跌宕起伏的连续创业[s6][s12]。他说自己“现在是从ICU往外走的路上”，小赢总比死好，但也不够[s12]。这位拥有69.3%投票权的创始人，正在将小鹏从一家汽车公司重塑为全球物理AI平台——汽车、飞行汽车、机器人构成三条增长曲线[s9]。他的财富故事同样惊人：胡润百富榜第114位、480亿元身家，造车新势力首富。但比财富更值得关注的是他的决策模式——砍掉花了“小几十个亿”的旧自动驾驶体系、拿掉公司名字里的“汽车”、把资源向机器人业务大幅倾斜，这些重塑公司命运的决定，本质上都是他一个人拍板的[s18]。从黄石到广州，从亚信到UC，从阿里到小鹏，从地面到天空——这位归来少年的征途，远未结束。

In the fourth quarter of 2025, XPeng posted its first-ever quarterly profit, of RMB 380 million, with its gross margin climbing to a record high of 21.3% [s7]. The story behind that number is far more dramatic than the financial statements suggest. It belongs to a programmer who came out of Huangshi, a small industrial city in Hubei province — a self-described "poor peasant" who moved offices six times in his first year as an entrepreneur, borrowed money every month just to make payroll, and could not even afford a plane ticket to pitch investors in Beijing [s3][s11]. From UC Browser, sold to Alibaba for US$4.35 billion, to building cars as one of "three outsiders," to the debacle of the G9 and the counterattack that followed; from intelligent driving to flying cars and on to humanoid robots — He Xiaopeng has spent two decades on one of the most turbulent entrepreneurial runs in the history of the Chinese internet [s6][s12]. "We are still on our way out of the ICU," he says. "A small win is better than death — but it is not enough" [s12]. The founder, who commands 69.3% of the company's voting rights, is now remaking XPeng from a carmaker into a global physical-AI platform, with cars, flying cars and robots forming three growth curves [s9]. His fortune is equally striking: ranked 114th on the Hurun Rich List with a net worth of RMB 48 billion, he is the wealthiest founder among China's new-energy-vehicle upstarts. But more noteworthy than the money is his decision-making style — scrapping an old autonomous-driving system that had cost "a few billion RMB," taking the word "automobile" out of the company's name, and tilting resources sharply toward the robotics business. Each of these company-reshaping calls was, in essence, made by one man [s18]. From Huangshi to Guangzhou, from AsiaInfo to UC, from Alibaba to XPeng, from the ground to the sky — the journey of the boy who came back is far from over.

## 正文 Archive Chapters

### 1. 早年求学：从黄石工人家庭到华南理工的程序员 / Early Years and Education: From a Worker's Family in Huangshi to a Programmer at South China University of Technology

1977年11月3日，何小鹏出生于湖北省黄石市一个普通工人家庭[s1]。父母都是普通职员，家庭条件并不宽裕，但给了他一个朴素而坚实的成长环境。祖籍洪湖市府场镇，湖楚大地的文化底蕴和务实精神在他身上留下了深深烙印。少年时代的何小鹏并不是那种传统意义上的'学霸'，但他对新鲜事物有着极强的好奇心和动手能力，尤其对电子和计算机表现出了超乎同龄人的兴趣。在那个年代，计算机对大多数家庭来说还是一个遥不可及的概念，但何小鹏的父母并没有因为经济条件的限制而忽视他的学习需求，这种开明的家庭氛围为他日后的技术之路打下了最初的基础。

1992年，何小鹏进入黄石二中求学[s1]。这所创建于1957年的省级重点中学以严谨的学风和较高的升学率闻名。在中学期间，他的理科成绩尤其突出，对数学和物理有着天然的亲近感。更重要的是，黄石二中当时已经开设了基础的计算机课程——虽然只是最简单的BASIC语言和打字练习——但这成为何小鹏接触信息技术的起点。他开始对'计算机能做什么'产生了朦胧的好奇，这种好奇在那个信息匮乏的年代显得尤为珍贵。

1995年，何小鹏考入华南理工大学计算机专业[s1][s2]。华南理工大学是当时中国工科教育的重镇之一，位于改革开放最前沿的广州。在大学期间，何小鹏展现出了典型的'技术型创业者'特质——他不仅系统地学习了计算机科学的理论和实践课程，更重要的是，他在广州这座商业气息浓厚的南方城市里，第一次真切地感受到了市场经济的活力和创业的可能性。

1997年暑假，何小鹏做了一件在当时看来相当超前的事——他和同学们把西18宿舍楼的所有电脑全部连上了教育网[s1]。在那个大多数人还在用拨号上网、网速慢得令人抓狂的年代，何小鹏和同学们用网线、集线器和简单的网络配置，把整栋宿舍楼的电脑连成了一个局域网，再接入校园网和Internet。这可能成为中国大学里最早实现整栋宿舍通网的案例之一。这件事不仅需要技术能力，更需要一种'先人半步'的行动力和组织能力——他需要说服楼管允许他们施工，需要协调同学们出人出力，还需要解决线材和设备的采购问题。

何小鹏后来总结这段经历时说了一句后来贯穿他整个创业生涯的话：'快人半步就是机会！'[s3]这句话看似简单，却蕴含着一个深刻的商业洞察：在技术变革的时代，最早行动的人往往能获得最大的先发优势——不是因为他们的技术最好，而是因为他们最早发现了需求并占据了用户心智。

1999年7月，何小鹏从华南理工大学毕业，获得计算机科学学士学位[s1][s2]。值得一提的是，他在2020年1月还获得了广东省人社厅颁发的工商管理高级经济师（科技型企业家）资格证书，这是对他后来创业成就的一种专业认可[s2]。

毕业后，何小鹏加入亚信科技（中国）有限公司广州分公司，历任技术经理、测试经理、项目经理[s1]。亚信科技是中国最早的互联网基础设施提供商之一，曾承建了中国电信Chinanet等六大运营商的互联网骨干网。何小鹏在这里参与了邮箱系统和企业级互联网解决方案的开发工作，从基层的技术岗位一步步做起。在亚信的四年多时间里，他不仅积累了丰富的技术管理和项目经验，也见证了中国互联网从萌芽到壮大的整个过程——从拨号上网到宽带接入，从门户网站到电子商务，从无人问津到全民上网。

2000年，亚信集团在纳斯达克上市，何小鹏获得了1500股原始股（市值约3万美元）[s1]。这是他人生中第一次感受到'资本'的力量——一家公司可以通过上市，让普通员工手中的股票变成真金白银。虽然3万美元在当时并不是天文数字，但这次经历让何小鹏对创业和资本之间的关系有了切身的理解，也为他日后创办UC优视埋下了种子。

在亚信期间，何小鹏的月薪远低于他的领导——他的领导月薪5万，而他的薪水不到领导的五分之一。这种差距让他感到'很难通过努力达到'财务自由的目标[s3]。但他并没有因此消极懈怠，而是在工作中不断积累经验、拓展人脉、观察市场。他后来曾半开玩笑地说，自己创业的动机之一，就是'出于贫下中农的嫉妒'[s3]——这种'不甘于被定义'的性格底色，为他日后的两次惊天创业埋下了种子。

**English:** He Xiaopeng was born on November 3, 1977, into an ordinary working-class family in Huangshi, Hubei province [s1]. Both of his parents were ordinary office workers; the household was far from well-off, but it gave him a plain yet solid environment in which to grow up. His ancestral home is Fuchang town in Honghu, and the cultural depth and pragmatic spirit of the lake country of Hubei left a deep imprint on him. The young He Xiaopeng was not a conventional "straight-A student," but he had an unusually strong curiosity about new things and a knack for making things with his hands, showing an interest in electronics and computers far beyond that of his peers. In those years a computer was still a remote concept for most families, yet his parents never let their limited means stand in the way of his appetite for learning. That open-minded family atmosphere laid the earliest foundation for the technological path he would later take.

In 1992, He Xiaopeng enrolled at Huangshi No. 2 High School [s1]. Founded in 1957, this provincial key school was renowned for its rigorous academic culture and its high university-admission rate. During his middle-school years he stood out particularly in the sciences, with a natural feel for mathematics and physics. More importantly, Huangshi No. 2 had already begun offering basic computer courses — nothing more than the simplest BASIC language and typing drills — but this became He Xiaopeng's first point of contact with information technology. He began to form a vague curiosity about "what a computer can do," a curiosity that was especially precious in that information-starved era.

In 1995, He Xiaopeng was admitted to South China University of Technology to study computer science [s1][s2]. SCUT was then one of the strongholds of Chinese engineering education, located in Guangzhou, the frontier city of reform and opening up. At university he displayed the classic traits of a "technical entrepreneur": not only did he systematically study the theoretical and practical curriculum of computer science, but in Guangzhou — a southern city thick with commercial energy — he felt, for the first time and at first hand, the vitality of the market economy and the possibility of starting a business.

In the summer of 1997, He Xiaopeng did something that seemed remarkably ahead of its time: he and his classmates wired every computer in the West Building 18 dormitory into the education and research network [s1]. In an age when most people were still torturing themselves with dial-up connections at maddeningly slow speeds, He and his classmates used network cables, hubs and simple network configurations to link the entire dormitory building's computers into a local area network, which they then connected to the campus network and the Internet. It may have been one of the earliest cases in a Chinese university of an entire dormitory building going online. The feat required not just technical ability but a "half-step-ahead" bias for action and a capacity for organization: he had to persuade the building manager to allow the work, coordinate classmates to contribute labor, and solve the procurement of cables and equipment.

Looking back on the episode, He Xiaopeng later summed it up in a sentence that would run through his entire entrepreneurial career: "Being half a step faster is the opportunity!" [s3] The line sounds simple, but it contains a deep business insight: in an era of technological change, the earliest movers usually reap the greatest first-mover advantage — not because their technology is the best, but because they are the first to spot a need and to occupy the user's mind.

In July 1999, He Xiaopeng graduated from South China University of Technology with a bachelor's degree in computer science [s1][s2]. It is worth noting that in January 2020 he also received a senior economist qualification certificate in business administration (technology-entrepreneur track) issued by the Guangdong Provincial Department of Human Resources and Social Security — a form of professional recognition for his later entrepreneurial achievements [s2].

After graduation, He Xiaopeng joined the Guangzhou branch of AsiaInfo Technologies (China) Co., Ltd., where he served successively as technical manager, testing manager and project manager [s1]. AsiaInfo was one of China's earliest internet-infrastructure providers, having built the internet backbone networks of six major operators, including China Telecom's Chinanet. There he took part in developing email systems and enterprise-grade internet solutions, working his way up step by step from junior technical posts. In his more than four years at AsiaInfo he not only accumulated rich experience in technical management and projects but also witnessed the entire arc of the Chinese internet from sprout to scale — from dial-up to broadband access, from portals to e-commerce, from indifference to mass adoption.

In 2000, AsiaInfo Group listed on Nasdaq, and He Xiaopeng received 1,500 founder shares worth roughly US$30,000 [s1]. It was the first time he felt the power of "capital": a company could, through a listing, turn the shares held by ordinary employees into hard cash. Thirty thousand dollars was no astronomical sum even then, but the experience gave him a first-hand understanding of the relationship between entrepreneurship and capital, and planted the seed that would later grow into UCWeb.

During his AsiaInfo years, He Xiaopeng's monthly salary was far below his boss's — the boss earned RMB 50,000 a month, while He made less than a fifth of that. The gap left him feeling that financial freedom was "very hard to reach through hard work" [s3]. Yet he did not grow passive or slack off; instead he kept accumulating experience at work, expanding his network and watching the market. He later joked, half in earnest, that one of his motives for going into business was "the envy of a poor peasant" [s3] — that underlying refusal to be defined by his circumstances planted the seeds of his two spectacular ventures to come.

### 2. UCWeb创业：从UCMail到移动浏览器之王 / The UCWeb Start-up: From UCMail to King of the Mobile Browser

2004年，27岁的何小鹏做出了人生中第一个重大决定——从亚信离职，与同在亚信的校友梁捷共同创办UC优视[s1][s3]。创业动机直白而朴素：'我创业的动机很简单，就是出于贫下中农的嫉妒。'他觉得每月拿几千块薪水无法实现财务自由，而他在亚信的领导月薪5万，他的薪水不到领导的五分之一，这种差距让他感到'很难通过努力达到'[s3]。这种直白的创业动机，在那个'为了改变世界'的互联网创业叙事中显得格外真实——何小鹏不是要改变世界，他只是想要改变自己的人生。

UC最初的方向是做移动邮件——UCMail。这与何小鹏和梁捷在亚信做邮箱系统的技术背景直接相关。他们觉得移动邮件是一个有前景的方向，毕竟企业级邮件系统是他们最擅长的领域。但市场并不买账。2004年的中国移动互联网还处于2G时代，绝大多数手机连WAP浏览器都用不利索，谁会需要用手机收发邮件？产品推出后反响平平，公司很快陷入了资金困境。

就在这个最艰难的时刻，命运安排了何小鹏创业路上的第一个贵人——网易创始人丁磊[s3]。丁磊听说搞亚信邮箱的人出来创业了，又听说他们资金困难，就约何小鹏出来喝酒。酒桌上的谈话直接而爽快，丁磊给何小鹏提供了80万元借款，还借给他网易的服务器，甚至在网易公司里给他开了一间办公室[s3]。丁磊为什么会帮助一个素不相识的年轻人？答案很简单——他也是技术出身，他理解这种创业者面临的困境。这80万对当时山穷水尽的UC来说无异于救命稻草。

拿到丁磊的支持后，何小鹏开始重新研究产品方向。转机出现在一个意想不到的地方——UCWeb浏览器最初其实是开发UCMail时的一个副产品[s3][s4]。他们在做邮件系统时需要处理WAP网站的访问问题，因为当时WAP网站加载极慢、流量极贵，他们顺手做了一个辅助浏览工具来压缩页面数据、加速网页加载。但何小鹏很快敏锐地意识到，这个'副产品'可能比邮件本身更有价值。

当时中国有大量手机用户使用的是功能机，上网体验极差。WAP网站最大的问题是内容非常少、上网速度慢、流量昂贵。而UC浏览器的云端加速技术恰好解决了这个痛点——它在服务器端把网页压缩后再发送到手机，大幅减少了流量消耗和加载时间。'快人半步就是机会！'何小鹏决定全面转向UC浏览器[s3]。

转型后的UC迎来了真正的地狱模式。创业第一年，UC搬了6次家，办公室都是借用朋友的场地。有时候一个晚上得自己扛着服务器从一个办公室搬到另一个办公室继续开发——服务器重达几十斤，几个年轻人趁着夜色偷偷搬进去，第二天早上再假装是正式入驻[s3][s4]。何小鹏每月出去借钱给员工发工资，同学、朋友、父母、兄弟都借遍了。那种'这个月借到了钱，下个月又不知道怎么办'的焦虑，几乎压垮了整个团队。

最令他记忆深刻的，是广州市天河区政府给UC发放了10万元中小企业科技扶持基金的那一天。拿到支票时何小鹏'一路不停地抖啊摇啊'，说：'我太高兴了！两个月不用出去借钱了。'[s3][s4]这10万元在后来UC被43.5亿美元收购的故事里显得微不足道，但对当时的何小鹏来说，那是实实在在的续命钱。

2006年，UC资金链再次濒临断裂。红杉资本考察UC时发生过一段令人心酸的故事——何小鹏和梁捷竟然说自己没钱买机票去北京陈述项目。这是红杉合伙人谈过的上千个项目里第一个要求投资人出钱买机票的公司，最终红杉也没投[s4]。这个细节生动地展现了UC当时有多穷——连去见投资人的路费都拿不出来。

经李学凌（YY创始人）牵线，何小鹏认识了时任联想投资副总裁的俞永福[s3][s4]。俞永福对UC项目很感兴趣，想以投资人身份入股，但联想投资没有通过这个项目。于是俞永福找上了金山总裁雷军。雷军对这个项目其实一知半解，他唯一的要求就是：'俞永福去我就投。'[s3]雷军看中的不是项目本身，而是人——他相信俞永福的判断力和执行力。2006年底，雷军以天使投资人身份给UC投了400万元[s3][s4]。俞永福随即辞去联想投资的职位，出任UC优视的CEO。

俞永福加入后做了三个关键决策，彻底改变了UC的命运[s4]：第一，通过'数据+理论'的方式得出结论：聚焦手机浏览器，阶段性放弃UCMail、手机阅读、手机社区等其他产品。这在当时是一个大胆的决定——砍掉其他产品意味着把所有赌注押在一个产品上。第二，拒绝了赚钱的企业业务诱惑，只做个人消费市场——'在初创公司里用中庸的方法非常危险'[s4]。企业业务来钱快但不性感，个人市场来钱慢但有爆发力。俞永福选择了后者。第三，融资时要求投资者要么投1000万美元，要么免谈——充足的资金能保证UC不会为融资分心，不为短期指标牺牲长远利益[s4]。

三人形成了UC的'三驾马车'[s3]：何小鹏想法多、比较理想主义；梁捷性格老成、擅研究技术；俞永福懂管理、懂资本、综合能力强。俞永福会帮何小鹏把天马行空的想法梳理成可执行的方案，他曾对着何小鹏说天津话：'你丫的，天天YY，让我们帮你梳理好，帮你去实现。'何小鹏和梁捷主动把CEO位置留空的做法本身就展现了他们的格局——'人都有弱点，也有专长'[s4]。

在战略聚焦下，UC浏览器迅速崛起。2007年8月，晨兴投资和联创策源联合投资1000万美元A轮[s1]。2009年6月，阿里巴巴集团对UC进行战略投资[s1]。到2012年，UC用户突破4亿，百度曾开价20亿美元收购被拒，俞永福说'UC是非卖品'[s5]。

2013年3月，阿里花费5.06亿美元增持UC，持股达66%[s5]。2013年8月，马云出任UC董事[s5]。这一切最终汇聚到2014年6月11日——阿里巴巴以约43.5亿美元全资收购UC优视，创下当时中国互联网史上最大并购纪录，远超百度收购91无线的19亿美元[s5][s6]。交易以股票为主、现金为辅，UC团队获得了1230万股阿里限制性股票，占阿里总股本的0.5%[s5]。俞永福进入了阿里8人战略决策委员会——这是阿里收购其他公司从未给过的待遇[s5][s6]。

**English:** In 2004, at the age of twenty-seven, He Xiaopeng made the first major decision of his life: he quit AsiaInfo and, together with Liang Jie, a fellow alumnus who also worked there, co-founded UCWeb [s1][s3]. His motive for starting a business was blunt and plain: "My reason for starting a business was very simple — it came from the envy of a poor peasant." He felt that a salary of a few thousand RMB a month could never deliver financial freedom; his boss at AsiaInfo earned RMB 50,000 a month while he made less than a fifth of that, a gap that left him feeling the goal was "very hard to reach through hard work" [s3]. Such a candid motive stood out as strikingly real amid the internet-era rhetoric of "changing the world" — He Xiaopeng was not out to change the world; he simply wanted to change his own life.

UC's initial direction was mobile email — UCMail. This grew directly out of the technical background He and Liang Jie had built doing email systems at AsiaInfo. They thought mobile email was a promising direction; after all, enterprise email systems were the field they knew best. But the market did not buy it. In 2004 China's mobile internet was still in the 2G era; the vast majority of phones could not even run a WAP browser smoothly, so who would need to send and receive email on a phone? The product drew a lukewarm response after launch, and the company quickly ran into financial trouble.

It was at this most difficult moment that fate delivered the first benefactor of He Xiaopeng's entrepreneurial career — Ding Lei, the founder of NetEase [s3]. Ding had heard that the people behind AsiaInfo's email system had gone out on their own, and then heard that they were short of money, so he invited He Xiaopeng out for drinks. The conversation over the table was direct and refreshing: Ding Lei lent He Xiaopeng RMB 800,000, lent him NetEase's servers, and even gave him an office inside NetEase's own company [s3]. Why would Ding help a young man he barely knew? The answer was simple — he too came from a technical background and understood the predicament such entrepreneurs faced. For UC, then at the end of its rope, that RMB 800,000 was nothing less than a lifeline.

With Ding Lei's support in hand, He Xiaopeng began rethinking the product direction. The turning point came from an unexpected quarter — the UCWeb browser was originally nothing more than a by-product of developing UCMail [s3][s4]. While building the email system they needed to handle access to WAP sites, because WAP pages then loaded extremely slowly and data traffic was extremely expensive, so they built a small auxiliary browsing tool to compress page data and speed up loading. He Xiaopeng was quick to sense that this "by-product" might be worth more than the email product itself.

At the time, huge numbers of Chinese mobile-phone users were on feature phones, and the internet experience on them was dreadful. The biggest problems with WAP sites were scarce content, slow speeds and costly data traffic. UC Browser's cloud-acceleration technology solved exactly that pain point: it compressed web pages on the server side before sending them to the phone, drastically cutting both data consumption and loading time. "Being half a step faster is the opportunity!" He Xiaopeng decided to pivot the whole company to UC Browser [s3].

After the pivot, UC entered a true hell mode. In its first year the company moved offices six times, borrowing space from friends. Some nights the founders themselves had to haul servers — each weighing dozens of kilograms — from one office to another to keep developing; a few young men would smuggle the machines in under cover of darkness and pretend the next morning that they had officially moved in [s3][s4]. Every month He Xiaopeng went out borrowing money to make payroll, tapping out classmates, friends, parents and brothers. The anxiety of "we raised the money this month, but who knows about next month" nearly crushed the entire team.

The moment he remembers most vividly was the day the Tianhe district government of Guangzhou awarded UC a RMB 100,000 science-and-technology support grant for small and medium enterprises. When he got the check, He Xiaopeng "couldn't stop shaking it and waving it all the way," saying: "I'm so happy! We won't have to go out borrowing money for two months" [s3][s4]. That RMB 100,000 would look trivial in the later story of UC's US$4.35 billion sale, but for He Xiaopeng at the time it was, quite literally, money that kept the company alive.

In 2006, UC's cash chain came to the brink of snapping again. When Sequoia Capital came to look at the company, a poignant episode unfolded: He Xiaopeng and Liang Jie actually said they could not afford plane tickets to Beijing to present their project. It was the first company in the thousand-plus projects the Sequoia partners had ever reviewed that asked the investors to pay for the airfare — and in the end Sequoia did not invest [s4]. The detail vividly shows how destitute UC was: it could not even scrape together the travel money to meet investors.

Through an introduction by Li Xueling (the founder of YY), He Xiaopeng met Yu Yongfu, then a vice president at Legend Capital [s3][s4]. Yu was keenly interested in UC and wanted to come in as an investor, but Legend Capital declined to approve the deal. So Yu took the project to Lei Jun, the president of Kingsoft. Lei actually understood only half of what the project was about; his one condition was: "If Yu Yongfu goes, I'll invest" [s3]. What Lei saw in the deal was not the project itself but the people — he trusted Yu Yongfu's judgment and ability to execute. At the end of 2006, Lei invested RMB 4 million in UC as an angel investor [s3][s4]. Yu Yongfu promptly resigned from Legend Capital and became UCWeb's CEO.

After joining, Yu Yongfu made three key decisions that completely changed UC's fate [s4]. First, using an approach of "data plus reasoning," he reached the conclusion that the company should focus on the mobile browser and shelve UCMail, mobile reading, mobile communities and its other products for the time being — a bold call at the time, since killing the other products meant betting everything on a single one. Second, he refused the tempting, cash-generating enterprise business and served only the individual consumer market: "In a start-up, using middling, middle-of-the-road methods is extremely dangerous" [s4]. Enterprise business paid quickly but was unsexy; the consumer market paid slowly but had explosive potential. Yu chose the latter. Third, when raising money he told investors it was US$10 million or no talk at all — ample capital would ensure UC was not distracted by fundraising and would not sacrifice long-term interests for short-term metrics [s4].

The three men formed UC's "troika" [s3]: He Xiaopeng brimmed with ideas and was something of an idealist; Liang Jie was steady and seasoned, strong in technology research; Yu Yongfu understood management and capital and was the most well-rounded. Yu would help He Xiaopeng turn his wild, sky-high ideas into executable plans; he once told He in his Tianjin dialect: "You, with your head in the clouds daydreaming every day — we're the ones who sort it all out and make it real for you." The fact that He Xiaopeng and Liang Jie had voluntarily left the CEO seat vacant itself spoke to their breadth of vision: "Everyone has weaknesses, and everyone has strengths" [s4].

With the strategy focused, UC Browser rose fast. In August 2007, Morningside Ventures and GGV Capital co-led a US$10 million Series A round [s1]. In June 2009, Alibaba Group made a strategic investment in UC [s1]. By 2012 UC's user base had broken through 400 million; Baidu offered US$2 billion to acquire it and was turned down, with Yu Yongfu declaring "UC is not for sale" [s5].

In March 2013, Alibaba spent US$506 million increasing its stake in UC to 66% [s5]. In August 2013, Jack Ma took a seat on UC's board [s5]. All of this converged on June 11, 2014, when Alibaba acquired UCWeb outright for approximately US$4.35 billion — setting a record at the time as the largest acquisition in Chinese internet history, far exceeding Baidu's US$1.9 billion purchase of 91 Wireless [s5][s6]. The deal was structured mainly in stock with a smaller cash component, and the UC team received 12.3 million restricted Alibaba shares, equal to 0.5% of Alibaba's total share capital [s5]. Yu Yongfu entered Alibaba's eight-person strategic decision-making committee — a seat Alibaba had never granted to an acquired company before [s5][s6].

### 3. 阿里岁月与中年危机：'UC只出现了一次' / The Alibaba Years and a Mid-life Crisis: 'UC Was Mentioned Just Once'

UC被阿里收购后，何小鹏出任阿里移动事业群总裁、阿里游戏董事长、土豆总裁[s1][s2]。从表面上看，这是一个创业者最好的归宿——公司被巨头高价收购，自己进入管理层，财务自由也得以实现。UC上上下下都是员工买房买车的消息，三位创始人还合资买了一艘价值几千万的游艇[s7]。按照正常的人生剧本，接下来的故事应该是享受人生、投资天使、偶尔上上综艺节目。但何小鹏不是按照正常剧本来的人。

在阿里当年的高管年会上，他从下午一直听到傍晚，在那些关于支付、电商、云计算的核心业务讨论中，他敏锐地意识到一件事——'UC'这个词从头到尾只出现了一次[s7]。自己做了十年的公司、43.5亿美元卖出去的UC，在整个阿里的战略版图里只被提了一个词。那种感觉就像一个父亲把自己辛苦养大的孩子卖给了一户大户人家，结果发现那户人家根本没把这孩子当回事。这种失落感是巨大的、难以言说的——不是对钱的不满，而是对自己过去十年付出的一种存在性怀疑。

何小鹏后来用'中年危机'来形容那个阶段。'逍遥子（张勇）就问，你多大，我说37不到，他说你这个年纪就中年危机了？我理解的中年危机是说我的梦想已经到这了，战斗的感觉没有了，我的下一步该如何走，不知道方向，也没有方向。'[s7]37岁的财务自由，带来的不是幸福，而是一种'拔剑四顾心茫然'的虚无。

在阿里的日子里，何小鹏负责的业务虽然体量不小，但并非阿里的核心战场。UC浏览器在移动互联网的浪潮中逐渐失去了此前的光芒——随着4G普及和原生APP生态崛起，浏览器作为上网入口的地位被大大削弱。微信、今日头条等超级APP的出现，让'浏览器'这个品类从'必装'变成了'可选'。土豆视频更是在与优酷合并后逐渐边缘化，何小鹏虽然身兼数职，却找不到UC时代那种'打仗'的兴奋感。

2014年6月13日，就在UC被正式收购两天后，特斯拉CEO马斯克宣布公司采用开源模式，对外开放所有专利[s1]。这个新闻深深触动了何小鹏。他在阿里内部建议做车，但并没有如愿推进下去——阿里的战略重心在电商和云计算，造车不在其规划之中[s7]。但造车的种子已经在他心里生根。同年，他联合YY创始人李学凌、猎豹移动CEO傅盛、腾讯高管吴宵光等人投资了一家名为'橙行智动'的互联网电动汽车公司——这就是小鹏汽车的前身[s1]。

何小鹏的角色最初只是天使投资人。他提供资金和资源，但没有打算亲自下场。真正促使他从投资人变成创业者的催化剂有两个。第一个是GGV纪源资本管理合伙人符绩勋的一通电话——'制造业升级，出行行业升级，已经大势所趋了。你要赶紧出来，再不出来这个窗口就没有了。'[s1]符绩勋是一个经验丰富的投资人，他的判断不是空穴来风——2014-2015年，全球范围内涌现了一大批智能电动车创业公司，特斯拉市值突破300亿美元，中国市场也出现了蔚来、比亚迪等先行者。窗口期确实存在，而且正在快速收窄。

第二个催化剂更加私人化。2017年2月16日，何小鹏的儿子出生。他说：'我有这个想法的时候是我儿子出生的第一个小时，我刚抱着他，接到一个电话……以前听了类似的话很多遍我都不为所动，但是那一天，那一刻，我突然想到我儿子长大了以后问我爸爸干什么工作的？我不想说退休了，所以对我很触动，决定要去做这样一件事。'[s7]他想做一些事情，让儿子将来问起'爸爸干什么的'时，能有一个值得骄傲的答案。'上一个十年可以告诉女儿爸爸干了一个很牛的事，下十年我希望能告诉儿子爸爸又做了一件什么事。'[s7]

2017年8月22日，UC十三岁生日当天，何小鹏正式从阿里离职。俞永福在内部信中说，何小鹏将正式'荣退'，并'开启自己新的一段追梦之旅'[s1]。那天，何小鹏发了一条朋友圈：'创业一轮回，苦辣酸甜咸，归来仍是少年。'[s1][s7]这14个字，浓缩了他对UC十年创业岁月的全部情感——有骄傲、有遗憾、有不舍、有释然。

值得称道的是，何小鹏离职时没有带走阿里部门的一个人——'要对老东家讲义气。'[s1]这种职业操守为他赢得了业界的广泛尊重，也为后来小鹏与阿里之间的合作保留了良好的关系基础。

何小鹏后来在接受采访时说：'我觉得中国赚到钱的一代创业者就应该多一点人出来去做再创业，做一些更大程度创造社会价值、创造全新意义的事情。因为他们有资源，包括钱、人脉和经验，能把一个很难的事情长时间地投入，做到成功。'[s7]这段话既是他的人生宣言，也是他对'连续创业者'这个标签的最好注解。

**English:** After UC was acquired by Alibaba, He Xiaopeng became president of Alibaba's Mobile Business Group, chairman of Alibaba Games and president of Tudou [s1][s2]. On the surface, this was the best possible outcome for an entrepreneur — his company bought at a high price by a giant, a seat in management, and financial freedom achieved. All through UC there was news of employees buying houses and cars, and the three founders even went in together on a yacht worth tens of millions of RMB [s7]. By the normal life script, the next chapter should have been enjoying life, angel investing and the occasional appearance on a variety show. But He Xiaopeng was not a man who followed the normal script.

At Alibaba's annual senior-executive conference that year, he sat listening from afternoon into evening. Amid the core-business discussions of payments, e-commerce and cloud computing, he noticed something sharply — the name "UC" appeared exactly once from beginning to end [s7]. The company he had built for ten years and sold for US$4.35 billion had been mentioned a single time across Alibaba's entire strategic map. It felt like a father who had sold the child he had painstakingly raised to a wealthy household, only to find that the household did not take the child seriously at all. The letdown was enormous and hard to articulate — not resentment over money, but an existential doubt about the value of the past decade of his life.

He Xiaopeng later described that phase as a "mid-life crisis." "Xiaoyaozi [Daniel Zhang] asked me how old I was. I said not yet thirty-seven. He said, 'You're having a mid-life crisis at this age?' What I understood a mid-life crisis to mean was that my dream had reached its limit, the will to fight was gone, and I had no idea where my next step should lead — no direction, and no sense of direction" [s7]. Financial freedom at thirty-seven brought not happiness but a kind of blank, sword-in-hand, looking-around-in-a-daze emptiness.

During his Alibaba days, the businesses He Xiaopeng ran were not small in scale, but they were not Alibaba's central battlefield either. UC Browser gradually lost its former sheen amid the mobile-internet tide: as 4G spread and the native-app ecosystem rose, the browser's position as the gateway to the internet was greatly weakened. The arrival of super-apps such as WeChat and Toutiao turned the browser category from a "must-install" into an "optional extra." Tudou, the video site, drifted further to the margins after its merger with Youku. Though he held several posts at once, He Xiaopeng could not find the thrill of "being in battle" that he had known in the UC era.

On June 13, 2014, just two days after UC's acquisition was formally completed, Tesla CEO Elon Musk announced that the company would adopt an open-source model and release all of its patents [s1]. The news struck He Xiaopeng deeply. Inside Alibaba he proposed building cars, but the idea did not move forward as he hoped — Alibaba's strategic center of gravity lay in e-commerce and cloud computing, and carmaking was not on its agenda [s7]. Yet the seed of carmaking had already taken root in him. That same year, together with Li Xueling (founder of YY), Fu Sheng (CEO of Cheetah Mobile), Wu Xiaoguang (a senior Tencent executive) and others, he invested in an internet electric-vehicle company called "Orange Xing Zhidong" — the predecessor of XPeng [s1].

He Xiaopeng's role at first was only that of an angel investor. He put up money and resources but had no intention of stepping into the arena himself. Two catalysts ultimately turned him from investor into entrepreneur. The first was a phone call from Hany Tan (Fu Jixun), managing partner of GGV Capital: "Manufacturing is upgrading, the mobility industry is upgrading — the trend is already set. You have to get out there fast; if you don't, this window will close" [s1]. Tan was a seasoned investor, and his judgment was no idle speculation: in 2014-15 a wave of smart-EV start-ups was emerging worldwide, Tesla's market value broke through US$30 billion, and the Chinese market had produced pioneers such as Nio and BYD. The window was real — and narrowing fast.

The second catalyst was more personal. On February 16, 2017, He Xiaopeng's son was born. He said: "I had this idea in the very first hour after my son was born. I was just holding him when I got a phone call... I had heard that kind of talk many times before and never been moved by it, but that day, at that moment, it suddenly occurred to me: when my son grows up and asks what his father does for a living, I don't want to say 'retired.' So it really touched me, and I decided I had to do something like this" [s7]. He wanted to do something that, when his son one day asked "what does Daddy do," would give him an answer worth being proud of. "In the last decade I could tell my daughter that Daddy did something pretty remarkable; in the next ten years I hope I can tell my son that Daddy did something remarkable all over again" [s7].

On August 22, 2017 — UC's thirteenth birthday — He Xiaopeng formally resigned from Alibaba. In an internal letter, Yu Yongfu wrote that He would formally "retire with honor" and "embark on a new journey in pursuit of his dreams" [s1]. That day He Xiaopeng posted to his WeChat Moments: "One full cycle of entrepreneurship, bitter and hot and sour and sweet and salty — the boy who comes back is still a boy" [s1][s7]. Fourteen characters distilling all his feeling for UC's ten-year run — pride, regret, reluctance and release.

Creditably, when He Xiaopeng left he did not take a single person from his Alibaba department with him: "You owe loyalty to your old employer" [s1]. That professional ethics won him broad respect across the industry and preserved a sound foundation of relations for the later cooperation between XPeng and Alibaba.

He Xiaopeng later said in an interview: "I think more of the generation of Chinese entrepreneurs who have made money should go out and start businesses again, doing things that create social value on a larger scale and create meaning that is entirely new. Because they have the resources — money, connections and experience — and can pour themselves into something hard for a long time and carry it through to success" [s7]. The passage is both his personal manifesto and the best footnote on the label "serial entrepreneur."

### 4. 小鹏汽车创立：三个外行与一个资本天团 / Founding XPeng: Three Outsiders and a Cast of Capital Heavyweights

和很多人以为的不同，小鹏汽车最初并不是何小鹏创立的。这家以他名字命名的公司，在诞生时与他没有直接关系[s8]。

2014年，三个从广汽集团出来的年轻人——夏珩（广汽研究院控制科科长）、何涛（广汽无人驾驶开发工程师）、杨春雷（丰田电动车项目经理）——在广州番禺大学城附近的一个车间里，手工组装汽车模型：他们把一台雷克萨斯SUV的发动机拆掉，换成电池[s8]。这家公司的名字叫'橙行智动'。三个年轻人都有扎实的汽车行业背景，但缺乏资金和商业资源。夏珩找到何小鹏时，何小鹏刚把UC卖给阿里实现财务自由，手里有钱也有资源，但自己完全不懂造车。他希望找到有经验的人来做创始人，自己提供资金就好。两人见面后一拍即合。

何小鹏随即张罗了一个'资本天团'[s8]：UC董事长俞永福、YY创始人李学凌、猎豹移动CEO傅盛、腾讯高管吴宵光等多位互联网大佬，共同出了一大笔天使投资。何小鹏自己也掏了约3亿元启动资金。这些互联网圈的朋友给了何小鹏最大的信任——他们不是汽车行业的专家，但他们相信何小鹏的判断力和执行力。在中国互联网圈，'跟何小鹏投'几乎成了一种共识——这个人要么不做，要做就一定能做到底。

公司本想取名'橙子汽车'，但没拿下商标。所有高管的名字都被拿去申请了一遍商标，都申请不下来，最后轮到何小鹏的名字才搞定，于是有了'小鹏汽车'这个名字[s8]。这个看似偶然的命名，日后成为中国家喻户晓的新能源汽车品牌。而何小鹏本人对这个以他名字命名的品牌有着一种复杂的感情——既是荣誉，也是压力。'小鹏'二字既是品牌名也是人名，产品的任何质量问题都会直接关联到他个人的声誉。

创业初期的艰难程度远超何小鹏的想象。连个正经办公室都没有，团队挤在广州天河区一间商住两用房里，白天画图改方案，晚上打地铺睡觉。去供应商那儿谈零件，人家一听是新成立的车企，门都不让进，说'骗子公司'[s8]。在那个年代，'互联网人造车'在汽车行业看来就是一个笑话——你懂什么叫冲压？懂什么叫焊接？懂什么叫NVH？你连一条生产线都没有，凭什么造车？

何小鹏后来回忆，那是他创业以来最绝望的时刻[s8]。资金快烧完了，投资人都躲着走，连员工都开始离职。他把自己的房子抵押了——在UC时代已经财务自由的人，又重新回到了'身无分文'的状态。他又拉来阿里巴巴和富士康投资，才凑够了续命的钱。这段经历与UC创业时期一年搬6次家的窘迫如出一辙——何小鹏似乎总能在最艰难的时刻找到活下去的办法。

2015年1月，小鹏汽车正式注册成立[s1]。2016年9月，Beta版正式发布[s1]。2017年8月29日，何小鹏正式加入出任董事长[s1]。此时公司已经运行了三年，有300人规模，第一款量产车呼之欲出。

何小鹏加入后做的第一个重大决定，就是推倒重做G3的Beta版。他认为现有版本达不到上市标准，需要重新打磨——这个决定让小鹏的第一款量产车上市推迟了近一年[s8]。在'总办'投票中，何小鹏的主张一度被否决，但他坚持己见，最终说服了团队。这个决定在当时看来是'多此一举'，但事后证明是正确的——G3虽然销量不算惊艳，但至少没有因为质量问题而翻车。

公司的最高决策机构是'总办'[s8]，由何小鹏（两票）、夏珩（一票）、何涛（一票）、顾宏地（一票）四人组成。但夏珩和何涛都是广汽出身、一起创业，意见几乎总是高度一致，这就形成了一种微妙的权力平衡。何小鹏虽然有兩票但并非绝对多数，这种'议会制'的决策模式在创业初期有利于集思广益，但在后来的市场竞争中暴露出效率不足的致命缺陷[s8]。

2018年12月，小鹏G3正式上市发布并启动交付[s1]。虽然G3的销量并不算惊艳，但它证明了小鹏汽车从PPT到量产的交付能力，也验证了互联网人造车的可行性。对何小鹏个人来说，G3的交付更像是一次自我证明——他不是'玩票'的，他是真的能造出车来的。

**English:** Contrary to what many assume, XPeng was not originally founded by He Xiaopeng. The company that bears his name had no direct connection with him at its birth [s8].

In 2014, three young men who had come out of GAC Group — Xia Heng (head of the control section at the GAC R&D Center), He Tao (a GAC autonomous-driving development engineer) and Yang Chunlei (a Toyota EV project manager) — were hand-assembling car models in a workshop near the university town in Panyu, Guangzhou: they had stripped the engine out of a Lexus SUV and replaced it with batteries [s8]. The company was called "Orange Xing Zhidong." All three had solid automotive backgrounds but lacked capital and business resources. When Xia Heng approached He Xiaopeng, He had just sold UC to Alibaba and achieved financial freedom; he had money and resources but knew nothing whatever about building cars. He hoped to find experienced people to serve as founders while he simply provided funding. The two met and clicked immediately.

He Xiaopeng promptly assembled a "cast of capital heavyweights" [s8]: UC chairman Yu Yongfu, YY founder Li Xueling, Cheetah Mobile CEO Fu Sheng, Tencent executive Wu Xiaoguang and several other internet luminaries collectively put up a large angel round. He himself also chipped in roughly RMB 300 million in start-up capital. His friends in the internet circle gave him their fullest trust — they were no experts in the auto industry, but they believed in He Xiaopeng's judgment and his ability to execute. In Chinese internet circles, "investing alongside He Xiaopeng" almost became a consensus: this is a man who either does nothing or sees a thing through to the end.

The company had wanted to call itself "Orange Auto," but could not secure the trademark. Every executive's name was tried in a trademark application and every one was rejected; when it finally came to He Xiaopeng's name, it went through — and so the company became "XPeng" [s8]. That seemingly accidental naming later grew into a new-energy vehicle brand known in every Chinese household. He himself has complicated feelings about the brand named after him — it is both an honor and a pressure. "Xiaopeng" is at once a brand name and a person's name, so any quality problem with a product attaches directly to his personal reputation.

The hardship of the early start-up far exceeded He Xiaopeng's imagination. There was not even a proper office; the team crowded into a combined residential-commercial apartment in Tianhe district, Guangzhou, drawing diagrams and revising plans by day and sleeping on the floor by night. When they went to suppliers to discuss parts, the suppliers, hearing it was a newly established car company, would not even let them through the door, calling them a "scam company" [s8]. In those days, "internet people building cars" was a joke as far as the auto industry was concerned: What do you know about stamping? What do you know about welding? What do you know about NVH? You don't even have a production line — what gives you the right to build cars?

He Xiaopeng later recalled that this was the most despairing moment of his entrepreneurial career [s8]. The cash was nearly burned through, investors were keeping their distance, and even employees were starting to leave. He mortgaged his own house — a man who had already achieved financial freedom in the UC era found himself back in a state of "pennilessness." He then brought in Alibaba and Foxconn as investors, and only then scraped together enough money to keep the company alive. The episode closely mirrored the desperation of the UC years, when the company moved offices six times in twelve months — He Xiaopeng seems always able to find a way to survive at the hardest moments.

In January 2015, XPeng was formally registered and incorporated [s1]. In September 2016, the Beta version was officially released [s1]. On August 29, 2017, He Xiaopeng formally joined as chairman [s1]. By then the company had been running for three years, had a headcount of 300, and its first mass-production car was close at hand.

The first major decision He Xiaopeng made after joining was to tear up and rebuild the Beta version of the G3. He judged that the existing version fell short of launch standard and needed to be reworked from scratch — a decision that pushed the launch of XPeng's first production car back by nearly a year [s8]. In a vote of the "general office," He's position was initially rejected, but he held his ground and eventually persuaded the team. The decision looked like a needless detour at the time, but it proved right in hindsight: while the G3's sales were unspectacular, it at least did not crash and burn over quality problems.

The company's highest decision-making body was the "general office" [s8], made up of four people: He Xiaopeng (two votes), Xia Heng (one vote), He Tao (one vote) and Brian Gu (Gu Hongdi, one vote). But Xia Heng and He Tao had both come out of GAC and started the business together, so their views were almost always closely aligned, creating a subtle balance of power. Although He Xiaopeng held two votes, he did not command an absolute majority. This "parliamentary" decision-making model helped pool wisdom in the start-up phase, but in the market competition that followed it exposed a fatal flaw: insufficient efficiency [s8].

In December 2018, the XPeng G3 was officially launched and deliveries began [s1]. Although the G3's sales were not spectacular, it proved XPeng's ability to deliver — to go from PowerPoint to mass production — and validated the feasibility of internet people building cars. For He Xiaopeng personally, the G3's delivery was more like a self-vindication: he was not dabbling; he really could build a car.

### 5. 资本之路与组织涅槃：从纽交所到'狠人'蜕变 / The Capital Road and an Organizational Rebirth: From the NYSE to the Making of a Tough Operator

小鹏汽车的融资之路，堪称中国造车新势力中最曲折的剧本之一。从2014年天使轮起步，经历了多轮融资，每一轮都伴随着市场质疑和资本博弈。造车是一个极度'烧钱'的行当——研发投入、工厂建设、供应链搭建、渠道铺设，每一个环节都需要天量资金。而小鹏作为一家'互联网人'创办的车企，在早期并没有获得传统汽车资本的足够信任。

2019年11月，小鹏完成C轮4亿美元融资，小米集团作为战略投资者入局[s1]。这笔融资的意义不仅在于资金本身——小米的入局意味着智能硬件生态与智能汽车的潜在协同，也给小鹏带来了品牌背书。但此时整个造车新势力赛道笼罩在'死亡阴影'之下：蔚来在2019年差点破产，威马汽车资金链紧张，资本市场对'造车新势力'五个字避之不及。小鹏能够拿到4亿美元C轮融资，某种程度上证明了投资者对何小鹏团队的长期信心。

2020年4月，小鹏P7上市[s9]。这款旗舰纯电轿车搭载最高706公里NEDC续航，直接对标特斯拉Model 3。P7的成功让小鹏在中高端纯电轿车市场站稳了脚跟，也让资本市场看到了小鹏的产品实力。2020年8月27日，小鹏汽车在纽约证券交易所正式上市，股票代码XPEV，发行价15美元/ADS，对应市值108.2亿美元，为当时全球新能源汽车最大IPO[s9]。敲钟的那一刻，何小鹏成为了中国造车新势力中最年轻的上市公司掌门人之一。

2021年7月7日，小鹏在香港联交所挂牌上市，成为首个在纽交所+港交所双重主要上市的中国造车新势力[s1]。双重上市不仅拓宽了融资渠道，也降低了中美资本市场脱钩的潜在风险——何小鹏在资本运作上的前瞻性再次得到体现。

然而，资本的青睐并不意味着一帆风顺。2022年9月，小鹏遭遇了创业以来最严峻的危机——G9上市失利[s1][s8]。这款被寄予厚望的旗舰SUV因三大致命问题遭遇市场冷遇：一是SKU配置极其复杂，消费者选车像做数学题；二是选配逻辑混乱，很多功能需要额外加钱；三是最终价格高于市场预期。发布会仅48小时内，小鹏被迫紧急改价[s8]。原定三个月3.5万辆的销售目标，实际季度交付仅6189辆——完成率不到18%[s8]。

G9失利成为何小鹏管理风格转变的转折点。此前他的治理风格被形容为'无为而治'[s8]。雷军曾直言不讳地说：'你最大的问题是你只想做董事长，不想做CEO。'[s8]这句话深深刺痛了何小鹏。在G9失利后的复盘中，他承认12位核心高管中只剩2位，公司组织效率出现了严重问题[s8]。

痛定思痛，何小鹏开启了大刀阔斧的组织变革。2023年1月30日，他邀请素有'车圈铁娘子'之称的王凤英正式加入出任总裁[s1]。王凤英在长城汽车工作31年，从销售成长为总经理，以铁腕手段和拼命精神著称[s8]。2023年全年，小鹏经历了'外科手术式'调整：一年更换10名高管，30个一级中心换了85%负责人；启动供应链反腐，采购负责人被调查[s8][s9]。

何小鹏自己也完成了从'老好人'到'狠人'的蜕变[s8][s9]。他说：'董事长可以做好人，但CEO不能。'[s9]他承认自己性格暴躁，办公室装了双层玻璃——一方面是为了隔绝外面王凤英的怒吼声，另一方面也是为了防止自己发火的声音影响员工工作[s8]。王凤英的加入不仅带来了管理上的铁腕手段，更重要的是帮助何小鹏完成了组织层面的'外科手术'——铲平山头，将过去创始团队的业务纳入何小鹏为主的新管理层手上[s8]。她是何小鹏从'老好人'走向'狠人'的关键推手。

**English:** XPeng's financing story is one of the most tortuous scripts among China's new-energy-vehicle upstarts. Starting from the angel round in 2014, it went through multiple rounds of financing, each accompanied by market skepticism and capital games. Carmaking is an extraordinarily cash-burning business — R&D spending, factory construction, supply-chain building and channel rollout all require astronomical sums at every stage. And XPeng, a car company founded by "internet people," did not enjoy enough trust from traditional automotive capital in its early years.

In November 2019, XPeng closed a US$400 million Series C round, with Xiaomi Group entering as a strategic investor [s1]. The significance of the round went beyond the money itself: Xiaomi's entry implied potential synergies between the smart-hardware ecosystem and smart cars, and it gave XPeng a brand endorsement. Yet the entire EV-upstart track then lay under a "shadow of death": Nio nearly went bankrupt in 2019, WM Motor's cash chain was under strain, and capital markets shunned the very phrase "new carmaking forces." That XPeng could land a US$400 million Series C round was, in a sense, proof of investors' long-term confidence in He Xiaopeng's team.

In April 2020, the XPeng P7 went on sale [s9]. The flagship all-electric sedan, with NEDC range of up to 706 kilometers, took aim directly at the Tesla Model 3. The P7's success anchored XPeng in the mid-to-high-end electric-sedan market and showed capital markets the company's product strength. On August 27, 2020, XPeng officially listed on the New York Stock Exchange under the ticker XPEV, at an offer price of US$15 per ADS, corresponding to a market value of US$10.82 billion — the largest IPO ever for a new-energy vehicle company globally at the time [s9]. At the moment of the bell-ringing, He Xiaopeng became one of the youngest listed-company chiefs among China's EV upstarts.

On July 7, 2021, XPeng listed on the Hong Kong Stock Exchange, becoming the first Chinese EV upstart to achieve dual primary listings on both the NYSE and the HKEX [s1]. Dual listing not only broadened financing channels but also reduced the potential risk of a decoupling between Chinese and American capital markets — He Xiaopeng's foresight in capital operations was in evidence once again.

Capital's favor, however, did not mean smooth sailing. In September 2022, XPeng met the gravest crisis of its history — the failed launch of the G9 [s1][s8]. The flagship SUV on which so much had been pinned met a cold reception for three fatal reasons: first, its SKU and configuration scheme was so complex that choosing a car felt like solving a math problem; second, the options logic was chaotic, with many features requiring extra payment; and third, the final prices came in above market expectations. Within just 48 hours of the launch event, XPeng was forced into an emergency repricing [s8]. Against an original target of 35,000 sales in three months, actual quarterly deliveries came to only 6,189 units — a completion rate below 18% [s8].

The G9 debacle became the turning point in He Xiaopeng's management style. Before it, his governance approach had been described as "laissez-faire" [s8]. Lei Jun had told him bluntly: "Your biggest problem is that you only want to be chairman, not CEO" [s8]. The words stung He Xiaopeng deeply. In the post-mortem after the G9 failure, he admitted that only two of the company's twelve core executives remained, and that the organization's efficiency had developed serious problems [s8].

Having learned the lesson the hard way, He Xiaopeng launched sweeping organizational change. On January 30, 2023, he invited Wang Fengying — known in the industry as the "iron lady of the car business" — to formally join as president [s1]. Wang had spent 31 years at Great Wall Motor, rising from sales to general manager, and was famed for her iron-handed methods and punishing work ethic [s8]. Through the whole of 2023, XPeng underwent "surgical" adjustments: in one year it replaced ten senior executives and changed 85% of the heads of its thirty first-level centers; it launched a supply-chain anti-corruption drive, and the procurement chief was placed under investigation [s8][s9].

He Xiaopeng himself completed the metamorphosis from "the nice guy" into "the tough operator" [s8][s9]. "A chairman can afford to play the good guy," he said, "but a CEO cannot" [s9]. He admits to a fiery temper, and his office was fitted with double-glazed glass — partly to shut out Wang Fengying's roaring from outside, and partly to keep the sound of his own outbursts from affecting his staff's work [s8]. Wang Fengying's arrival brought not only iron-handed management but, more importantly, helped He Xiaopeng perform the organizational "surgery" — flattening fiefdoms and bringing the businesses of the old founding team under the new, He-led management [s8]. She was the decisive catalyst in his journey from nice guy to tough operator.

### 6. 产品矩阵：从G3到MONA的十年造车路 / The Product Matrix: A Decade of Carmaking from the G3 to MONA

小鹏汽车的产品迭代史，是一部中国造车新势力从'跟随者'到'定义者'的成长史。从2018年G3的首次交付，到2026年MONA系列全面覆盖10-15万元市场，小鹏用十年时间完成了完整的产品矩阵布局——从紧凑型SUV到旗舰轿车，从纯电MPV到增程车型，从20万级中高端到10万级大众化市场。

2018年12月，小鹏G3正式上市并启动交付[s1]。这款紧凑型纯电SUV定位年轻消费群体，补贴后售价13-16万元区间，搭载L2级自动驾驶辅助功能。虽然销量不算惊艳，但G3证明了小鹏具备从PPT到量产的交付能力。

2020年4月，小鹏P7上市[s9]。这是小鹏的'破局之作'——旗舰纯电轿车，NEDC续航最高706公里，直接对标特斯拉Model 3。P7搭载了XPILOT 3.0智驾系统，配备14个摄像头、5个毫米波雷达及12个超声波传感器。P7的成功让小鹏在中高端纯电轿车市场站稳了脚跟，月销量一度突破7000台。

2021年9月，小鹏P5正式上市——全球首款搭载量产激光雷达的智能汽车[s1]。P5搭载了两个大疆览沃生产的激光雷达，标志着小鹏在感知硬件上的前瞻性布局。虽然P5的市场表现并不突出，但激光雷达的率先量产为后来XNGP城市智驾能力的落地奠定了硬件基础。

2022年9月，小鹏G9首次上市但因定价失利[s1]。这次失败成为小鹏历史上最大的产品教训——复杂的SKU配置和混乱的定价策略让消费者无所适从。但失败也带来了重生，G9后来经历了彻底的改款，重新定义了产品逻辑，简化了配置选择，聚焦核心卖点。

2023年4月，SEPA 2.0扶摇架构正式发布，小鹏G6全球首秀[s1]。扶摇架构是小鹏的第二代整车平台，支持800V高压SiC平台、CIB电池一体化等技术，大幅降低了制造成本。G6于2023年6月正式上市，成为小鹏走出G9阴影的关键车型。

2024年1月，小鹏X9正式上市——旗舰纯电MPV[s1]。X9瞄准家庭出行和高端商务场景，搭载后轮转向、空气悬架等高端配置，展现了小鹏在多样化车型上的技术实力。

2024年8月，MONA M03正式上市，起售价11.98万元[s9]。这是小鹏汽车的一个里程碑时刻。MONA品牌源自滴滴内部'达芬奇'造车项目，2023年8月小鹏以最高58.35亿港元股份对价收购[s10]。MONA主打10-15万元价格区间的年轻用户市场，M03上市52分钟大定破万，48小时超3万辆[s9]。截至2025年11月累计交付超20万辆[s1]。2025年全年MONA品牌交付17.53万辆，占总交付超40%[s9]。

2024年11月P7+上市[s1]，2026年1月8日一口气发布四款新车含首款增程G7[s1]。6月22日MONA L03 SUV发布[s9]。增程车型的推出意味着小鹏开始拥抱更广泛的用户需求——纯电和增程双路线并行。

从产品战略角度看，何小鹏在十年间完成了三个关键布局：第一，通过P7、G9、X9建立中高端品牌形象；第二，通过MONA品牌实现价格下探，覆盖最大众化的市场；第三，通过增程+纯电双动力路线，满足不同用户的补能焦虑。这套产品矩阵的核心逻辑是'技术向下'——用高端车型验证的技术能力，通过平台化和规模化，下沉到更大规模的市场。

**English:** XPeng's product-iteration history is a growth story of a Chinese EV upstart moving from "follower" to "definer." From the first deliveries of the G3 in 2018 to the full coverage of the RMB 100,000-150,000 market by the MONA line in 2026, XPeng spent a decade completing its product-matrix layout — from compact SUV to flagship sedan, from all-electric MPV to range-extended models, from the RMB 200,000-tier mid-to-high end to the RMB 100,000-tier mass market.

In December 2018, the XPeng G3 was officially launched and deliveries began [s1]. Aimed at young consumers, this compact all-electric SUV was priced in the RMB 130,000-160,000 range after subsidies and came with L2-level driver-assistance features. Its sales were not spectacular, but the G3 proved XPeng's ability to deliver from PowerPoint to mass production.

In April 2020, the XPeng P7 went on sale [s9]. This was XPeng's breakout model — a flagship all-electric sedan with NEDC range of up to 706 kilometers, positioned directly against the Tesla Model 3. The P7 carried the XPILOT 3.0 intelligent-driving system, equipped with fourteen cameras, five millimeter-wave radars and twelve ultrasonic sensors. Its success anchored XPeng in the mid-to-high-end electric-sedan market, with monthly sales at one point breaking 7,000 units.

In September 2021, the XPeng P5 was officially launched — the world's first mass-produced smart car fitted with lidar [s1]. The P5 carried two Livox lidar units made by DJI, signaling XPeng's forward-looking bet on perception hardware. The P5's market performance was not outstanding, but its early mass-production of lidar laid the hardware foundation for the later rollout of XNGP urban intelligent-driving capabilities.

In September 2022, the XPeng G9 was first launched but stumbled over pricing [s1]. The failure became the costliest product lesson in XPeng's history — its bewildering SKU complexity and chaotic pricing left consumers at a loss. But failure also bred rebirth: the G9 later underwent a thorough redesign, redefining the product logic, simplifying configuration choices and focusing on core selling points.

In April 2023, the SEPA 2.0 "Fuyao" architecture was officially released, and the XPeng G6 made its global debut [s1]. The Fuyao architecture was XPeng's second-generation vehicle platform, supporting an 800V high-voltage SiC platform, cell-integrated-body (CIB) battery technology and more, sharply reducing manufacturing costs. The G6 went on sale in June 2023 and became the pivotal model in XPeng's emergence from the shadow of the G9.

In January 2024, the XPeng X9 was officially launched — a flagship all-electric MPV [s1]. Aimed at family travel and high-end business use, the X9 carried rear-wheel steering, air suspension and other premium features, demonstrating XPeng's technical strength across diverse body styles.

In August 2024, the MONA M03 went on sale with a starting price of RMB 119,800 [s9]. It was a milestone moment for XPeng. The MONA brand originated from Didi's in-house "Da Vinci" carmaking project, which XPeng acquired in August 2023 for share consideration of up to HK$5.835 billion [s10]. MONA targets young users in the RMB 100,000-150,000 price band; the M03 topped 10,000 firm orders within 52 minutes of launch and 30,000 within 48 hours [s9]. Cumulative deliveries exceeded 200,000 units by November 2025 [s1]. For full-year 2025 the MONA brand delivered 175,300 units, accounting for over 40% of total deliveries [s9].

The P7+ went on sale in November 2024 [s1], and on January 8, 2026, four new models were launched at once, including the first range-extended model, the G7 [s1]. On June 22 the MONA L03 SUV was unveiled [s9]. The introduction of range-extended vehicles signaled XPeng's embrace of broader user needs — running pure-electric and range-extended routes in parallel.

From a product-strategy perspective, He Xiaopeng completed three key layouts over the decade: first, building a mid-to-high-end brand image through the P7, G9 and X9; second, reaching downward in price through the MONA brand to cover the largest mass market; and third, meeting different users' range-anxiety needs through the dual powertrain routes of range-extension and pure electric. The core logic of this product matrix is "technology flowing downward" — taking the technical capabilities validated on high-end models and, through platformization and scale, pushing them down into a much larger market.

### 7. 智驾之战：从XNGP到图灵AI大模型 / The Intelligent-Driving War: From XNGP to the Turing AI Large Model

智能驾驶是小鹏汽车最核心的技术标签，也是何小鹏投入最坚定、争议最大的领域。从早期的XPILOT到如今的图灵AI智驾（第二代VLA），小鹏走过了一条从规则驱动到数据驱动、从人工编码到AI大模型的技术进化之路。这条路线的本质，是让汽车从'被编程的机器'变成'能学习的AI'。

2021年9月，小鹏P5成为全球首款量产搭载激光雷达的智能汽车[s1]。两个大疆览沃激光雷达的配备，为小鹏的智驾系统提供了强大的感知冗余。2022年9月，城市NGP在广州行业首发——这是中国首个面向城市复杂道路场景的导航辅助驾驶功能[s1]。这意味着小鹏的智驾系统开始从高速公路延伸到城市道路，面对红绿灯、行人、非机动车等更加复杂的场景。

XNGP是小鹏智驾系统的集大成者，覆盖高速、城市、泊车全场景。到2025年底，XNGP已在全国范围内开通，无需高精地图即可在全国各地的城市道路上使用。这与特斯拉FSD的'纯视觉'路线形成了有趣的技术分野——小鹏选择'视觉+激光雷达'的融合感知方案，在保证安全冗余的同时逐步向无图方向演进。

小鹏与特斯拉FSD的技术路线之争，是近年来智能驾驶领域最受关注的话题之一[s11]。特斯拉坚持纯视觉方案（摄像头+神经网络），不依赖激光雷达和高精地图；小鹏则采用融合感知方案，但也在逐步减少对外部传感器的依赖。两者都在追求让车辆像人类一样'看懂'道路。

根据媒体实测数据，在800公里复杂路况的对比测试中，特斯拉FSD接管15次，华为ADS 2.0接管2次，小鹏XNGP接管6次[s11]。小鹏智驾副总裁刘先明曾表示，小鹏图灵AI智驾'领先行业5倍'[s11]。这一说法虽然引发了争议，但也反映了小鹏在智驾领域的强烈自信。

2026年3月，小鹏图灵AI智驾（第二代VLA）正式推送[s9]。这是小鹏智驾技术的又一次代际飞跃。VLA（Vision-Language-Action）大模型的核心参数令人瞩目：720亿参数、近1亿个视频训练片段、每5天全链路迭代一次[s11]。车端搭载小鹏自研的图灵芯片，配合端侧编译器，推理效率提升12倍[s11]。

第二代VLA的核心能力包括：30秒记忆陌生路线并自主行驶、6秒预判复杂交通场景、无需高精地图即可在全国任意道路上行驶[s11]。这些数据背后是一个庞大的数据飞轮——每一辆在路上行驶的小鹏汽车都在为AI大模型提供训练数据，数据越多模型越强，体验越好，用户越多，数据也越多。这种'数据飞轮'效应的建立，意味着后来者想要追赶小鹏的智驾水平，不仅需要同等级的技术投入，还需要时间来积累足够的数据量。

何小鹏对智驾技术的投入近乎偏执。他做了一个令很多人震惊的决定——把公司花了几年时间、耗资'小几十个亿'的旧自动驾驶体系全部砍掉，押注第二代VLA大模型路线[s8]。当时每个月要投入3个亿。这些决定都是他一个人拍板的——如果VLA路线最终被证明是错误的，那数十亿的投入将全部打水漂。但他愿意承担这个风险，因为他坚信'数据驱动'才是智能驾驶的正确方向。

何小鹏在2026年5月的一次对话中提出了一个引人注目的概念：'新的十年是两种新能源的叠加。第一种是过去所说的新能源——电能替代石油。而第二种新的能源，就是数据能源……原来数据是下一代汽车的重要新燃料，而芯片和模型是新的引擎。'[s12]这个'数据即燃料'的比喻，精准地概括了智能驾驶时代的竞争逻辑——谁拥有更多高质量的驾驶数据，谁就拥有更强的AI能力。

**English:** Intelligent driving is XPeng's most core technical label, and the field into which He Xiaopeng has invested most resolutely and which has aroused the most controversy. From the early XPILOT system to today's Turing AI driving (the second-generation VLA), XPeng has traveled an evolutionary path from rule-driven to data-driven, from hand-written code to large AI models. The essence of this route is to turn the car from "a programmed machine" into "an AI that can learn."

In September 2021, the XPeng P5 became the world's first mass-produced smart car equipped with lidar [s1]. The two DJI Livox lidar units gave XPeng's driving system powerful perception redundancy. In September 2022, City NGP made its industry debut in Guangzhou — China's first navigation-guided driver-assistance function for complex urban road scenarios [s1]. It meant XPeng's system was beginning to extend from highways into city streets, confronting far more complex scenes: traffic lights, pedestrians and non-motorized vehicles.

XNGP is the culmination of XPeng's intelligent-driving system, covering the full range of highway, city and parking scenarios. By the end of 2025, XNGP had been opened nationwide, usable on urban roads across the country without HD maps. This forms an intriguing technical divergence from Tesla FSD's "pure vision" route — XPeng chose a fused perception scheme of "vision plus lidar," preserving safety redundancy while gradually evolving toward a map-free approach.

The contest over technical routes between XPeng and Tesla FSD has been one of the most closely watched topics in intelligent driving in recent years [s11]. Tesla holds to a pure-vision approach (cameras plus neural networks), relying on neither lidar nor HD maps; XPeng uses a fused-perception approach but is likewise gradually reducing its dependence on external sensors. Both are pursuing the goal of making vehicles "understand" the road the way humans do.

According to media road-test data, in a comparative test over 800 kilometers of complex road conditions, Tesla FSD required 15 driver takeovers, Huawei ADS 2.0 required 2, and XPeng XNGP required 6 [s11]. Liu Xianming, XPeng's vice president of intelligent driving, has said that XPeng's Turing AI driving is "five times ahead of the industry" [s11]. The claim sparked controversy, but it reflects XPeng's fierce confidence in its driving technology.

In March 2026, XPeng's Turing AI driving (second-generation VLA) was officially pushed to vehicles [s9]. It marked another generational leap in XPeng's driving technology. The core parameters of the VLA (Vision-Language-Action) model are striking: 72 billion parameters, nearly 100 million video training clips, and a full-stack iteration once every five days [s11]. On the vehicle runs XPeng's self-developed Turing chip which, paired with an on-device compiler, lifts inference efficiency twelvefold [s11].

The second-generation VLA's core capabilities include memorizing an unfamiliar route in 30 seconds and then driving it autonomously, anticipating complex traffic scenes six seconds ahead, and driving on any road in the country without HD maps [s11]. Behind these figures lies a vast data flywheel: every XPeng car on the road feeds training data back to the AI large model — the more data, the stronger the model; the better the experience, the more users; and the more users, the more data. The establishment of this "data flywheel" effect means that a latecomer hoping to catch XPeng's driving level needs not only comparable technical investment but also time to accumulate a sufficient volume of data.

He Xiaopeng's commitment to driving technology borders on the obsessive. He made a decision that shocked many: he scrapped the company's entire old autonomous-driving system — years of work that had cost "a few billion RMB" — and bet everything on the second-generation VLA large-model route [s8]. At the time the program was consuming RMB 300 million a month. These decisions were all made by him alone: if the VLA route ultimately proved wrong, billions in investment would be entirely wasted. But he was willing to take that risk, because he firmly believed that "data-driven" was the correct direction for intelligent driving.

In a conversation in May 2026, He Xiaopeng put forward a striking concept: "The new decade is the superposition of two kinds of new energy. The first is the new energy we used to talk about — electricity replacing oil. The second new energy is data energy... It turns out data is the important new fuel for the next generation of cars, while chips and models are the new engines" [s12]. This metaphor of "data as fuel" captures precisely the competitive logic of the intelligent-driving age: whoever holds more high-quality driving data holds the stronger AI capability.

### 8. 大众合作：从技术引进到技术输出的历史性转折 / The Volkswagen Partnership: A Historic Turn from Technology Import to Technology Export

2023年7月，一则震惊全球汽车行业的消息传出：大众汽车集团宣布向小鹏汽车增资约7亿美元，持有4.99%股权，双方将联合开发两款B级电动车[s9]。这笔交易的深层意义远超财务投资——小鹏将向大众输出CEA（China E-Architecture）电子电气架构技术[s9]。

在汽车行业的百年历史中，通常是外资车企向中国合资企业提供技术平台和市场经验。德系车企输出底盘调校、发动机技术、品控体系；日系车企输出精益生产和混动技术。中国车企长期扮演'学生'角色，用市场换技术，用利润换经验。而这一次，方向完全反了过来——一家成立不到十年的中国新能源车企，向全球领先汽车集团之一输出核心电子架构[s9]。

合作的达成并非一蹴而就。大众方面最初是通过行业渠道了解到小鹏在电子电气架构方面的技术积累。CEA架构是小鹏在SEPA 2.0扶摇平台上开发的核心技术，集成了中央计算+区域控制的电子电气架构，具有高度灵活性和可扩展性，能够支持不同动力形式（纯电、插混、燃油）和不同级别车型的电子系统需求[s9]。

谈判过程中，大众方面对小鹏的技术能力进行了严格的评估和验证。大众作为全球车企的标杆，其技术标准的严苛程度可想而知——德国工程师对技术细节的要求是出了名的苛刻。小鹏能够通过与大众的技术合作评审，本身就是对其技术实力的一种权威认可。

合作很快取得了实质性进展。双方联合开发的两款B级电动车已进入实质性阶段，合作范围从最初的纯电平台扩展到燃油和插混平台[s9]。这意味着大众不仅看中了小鹏在纯电领域的技术，也认可其在动力总成多元化方面的能力。合作从'学一个产品'升级为'学一套体系'。

技术合作的财务回报立竿见影。2025年第四季度，小鹏服务及其他收入达到31.8亿元，同比增长121.9%，毛利率高达70.8%[s10]。这其中，大众的技术授权收入是重要组成部分。该季度贡献毛利约22.5亿元，占全公司总毛利近一半。技术授权本质上是一种高毛利的'卖水'生意——前期研发投入已经沉淀，后续的授权边际成本极低。

2026年4月，双方合作进一步深化，大众和小鹏宣布将联合开发更多车型，并探索在软件定义汽车领域的合作[s9]。有媒体分析，技术授权每季度可为小鹏带来约10亿元收入——虽然这只是媒体预测而非公司官方指引，但反映了市场对这一合作商业价值的高度期待。

对何小鹏个人而言，大众合作的意义同样重大。这证明了他一直坚信的'技术至上'路线是正确的——小鹏不是靠价格战和营销噱头取胜的车企，而是真正拥有核心技术能力的科技企业。他在多次公开场合强调，小鹏的目标是成为一家'技术驱动型'公司。大众合作就是这种技术驱动理念最好的背书——当全球最挑剔的汽车集团愿意为你的技术买单时，技术实力的含金量不言而喻。

大众合作的示范效应也辐射到了整个中国新能源汽车产业。它向全球汽车市场传递了一个明确信号：中国新能源车企不仅在市场规模上领先，在核心技术上也开始具备向全球主流车企输出技术的能力。这种技术输出的商业模式——研发投入→技术沉淀→对外授权→持续收入——为中国车企探索出了除卖车之外的第二条盈利路径。

**English:** In July 2023, news broke that stunned the global auto industry: Volkswagen Group announced a roughly US$700 million capital injection into XPeng, taking a 4.99% equity stake, and the two sides would jointly develop two B-class electric vehicles [s9]. The deeper significance of the deal far exceeded a financial investment — XPeng would export its CEA (China E-Architecture) electrical and electronic architecture technology to Volkswagen [s9].

In the century-long history of the auto industry, it has usually been foreign carmakers that supplied technology platforms and market know-how to Chinese joint ventures. German manufacturers exported chassis tuning, engine technology and quality-control systems; Japanese manufacturers exported lean production and hybrid technology. Chinese carmakers long played the role of "student," trading market access for technology and profits for experience. This time, the direction was completely reversed — a Chinese new-energy carmaker founded less than a decade earlier was exporting core electronic architecture to one of the world's largest auto groups [s9].

The partnership did not come about overnight. Volkswagen initially learned of XPeng's accumulated expertise in electrical and electronic architecture through industry channels. The CEA architecture was the core technology XPeng had developed on its SEPA 2.0 Fuyao platform, integrating a centralized-computing-plus-zonal-control electrical/electronic architecture with a high degree of flexibility and scalability, able to support the electronic-system needs of different powertrain forms (pure electric, plug-in hybrid, combustion) and different vehicle classes [s9].

During the negotiations, Volkswagen subjected XPeng's technical capabilities to rigorous assessment and verification. As the benchmark of the global auto industry, Volkswagen's technical standards were, predictably, unsparing — German engineers are famously exacting about technical detail. That XPeng passed the technical-cooperation review with Volkswagen was in itself an authoritative recognition of its technical strength.

The cooperation quickly produced substantive progress. The two B-class electric vehicles being jointly developed have entered the substantive stage, and the scope of cooperation has expanded from the original pure-electric platform to combustion and plug-in-hybrid platforms [s9]. This means Volkswagen values not only XPeng's technology in pure electrics but also recognizes its capability across diversified powertrains. The cooperation has upgraded from "learning one product" to "learning an entire system."

The financial returns from the technology cooperation were immediate. In the fourth quarter of 2025, XPeng's services and other revenues reached RMB 3.18 billion, up 121.9% year on year, with a gross margin as high as 70.8% [s10]. Of this, the Volkswagen technology-licensing income was a major component, contributing roughly RMB 2.25 billion in gross profit that quarter — nearly half of the company's total gross profit. Technology licensing is, in essence, a high-margin "selling shovels" business: the upfront R&D investment has already been sunk, and the marginal cost of subsequent licenses is extremely low.

In April 2026, the two sides deepened the partnership further: Volkswagen and XPeng announced they would jointly develop more models and explore cooperation in the field of software-defined vehicles [s9]. Some media have analyzed that technology licensing could bring XPeng around RMB 1 billion in revenue each quarter — though this is a media forecast rather than official company guidance, it reflects the high expectations the market holds for the commercial value of the cooperation.

For He Xiaopeng personally, the Volkswagen partnership matters just as much. It vindicates the "technology first" route in which he has always believed — XPeng is not a carmaker that wins on price wars and marketing gimmicks, but a technology company that genuinely owns core capabilities. He has stressed on many public occasions that XPeng's aim is to become a "technology-driven" company. The Volkswagen deal is the best possible endorsement of that technology-driven philosophy: when the world's most demanding auto group is willing to pay for your technology, the gold content of that technical strength speaks for itself.

The demonstration effect of the Volkswagen partnership has radiated across the entire Chinese new-energy-vehicle industry. It has sent a clear signal to the global auto market: Chinese new-energy carmakers lead not only in market scale but are beginning to possess the capability to export technology to the world's mainstream carmakers. This business model of technology export — R&D investment, technology accumulation, outward licensing, sustained income — has opened up a second path to profitability for Chinese carmakers beyond selling cars.

### 9. 飞行汽车与机器人：物理AI的三条曲线 / Flying Cars and Robots: The Three Curves of Physical AI

2022年9月，小鹏G9首次上市，成为小鹏汽车历史上最惨痛的一次产品发布。问题出在三个方面：一是SKU配置极其复杂，消费者选车像做数学题；二是选配逻辑混乱，很多功能需要额外加钱；三是最终价格高于市场预期。发布会48小时内，小鹏被迫紧急宣布降价增配、简化配置表，但市场已经不买账了[s6]。G9被寄予厚望，原计划三个月冲上3.5万辆，结果一个季度总交付量只有6189辆[s6]。更可怕的是，整个小鹏的销量都被带崩了——从2022上半年的近7万台，下滑到下半年的两万多，被均价更高的理想和蔚来按在地上摩擦。

背后的制度症结是“议会制”决策的失败。G9版本设定上，何小鹏原本希望不设置低智能化、低续航的入门款，聚焦中长续航，但最终没能说服其他高管，推出了多版本[s6]。这暴露了早期治理模式的根本问题——决策效率低、责任不清晰。公司的最高决策机构是“总办”，由何小鹏（两票）、夏珩（一票）、何涛（一票）、顾宏地（一票）四人组成。但夏珩和何涛都是广汽出身、一起创业，意见几乎总是高度一致。这就形成了“朝小野大”的局面[s6]。

9月末的一次高管会上，气氛凝重，何小鹏痛陈G9发布的种种问题。雷军曾直言不讳地指出小鹏最大的问题——“你只想做董事长，不想做CEO”。这句话成为何小鹏管理风格转变的转折点[s12]。G9失利后，何小鹏开始亲自管研发、采购、横向流程，管得最细的是采购[s12]。

G9失利后，何小鹏开始了大刀阔斧的组织变革。他取消了总裁办，建立战略、产规、技术规划、产销平衡、OTA五大虚拟委员会组织和E、F、H三大平台产品矩阵，三大平台负责人均直接向何小鹏汇报，权力开始向他一人集中。夏珩于2022年底辞任执行董事[s6]。2023年1月30日，素有“车圈铁娘子”之称的王凤英正式加入小鹏汽车出任总裁[s6]。

2023年，人事震荡进一步加剧。那一年，小鹏汽车30个一级中心的负责人换了85%，第二季度员工离职率飙升至18%。年底的G9二次发布会上，何小鹏坦言：“去年小鹏财报里的12位高管，今年只剩下两位。”[s12]王凤英入职前便敏锐察觉小鹏的钢材采购存在问题，何小鹏亲自带队核查，最终揪出供应链腐败链条——采购部负责人兼总裁李丰被警察带走，退休的制造副总裁蒋平也受到牵连。这既是反腐，也是权力重构[s6]。

何小鹏在2025年初接受晚点LatePost专访时说：“现在是从ICU往外走的路上……我们今天只是小赢，小赢总比死好，但也不够。”[s12]这句话浓缩了G9失利后两年多时间的挣扎与坚持。办公室装上双层玻璃——一半是隔绝外面王凤英的怒吼，一半是防止自己发火的声音传出去[s12]。“董事长可以做好人，但CEO不能。”[s12]这句话成为了小鹏组织革命的精神注脚。

**English:** In September 2022, the XPeng G9 was launched for the first time, in what became the most painful product launch in XPeng's history. The problems lay in three areas: first, the SKU and configuration scheme was so complex that choosing a car felt like solving a math problem; second, the options logic was chaotic, with many features requiring extra payment; and third, the final prices came in above market expectations. Within 48 hours of the launch event, XPeng was forced into an emergency announcement of price cuts, added features and a simplified configuration sheet, but the market had already stopped buying [s6]. The G9 had been expected to charge to 35,000 units in three months; instead, total deliveries for the whole quarter came to only 6,189 vehicles [s6]. Worse still, XPeng's overall sales were dragged down with it — from nearly 70,000 units in the first half of 2022 to just over 20,000 in the second half, as the company was ground into the floor by Li Auto and Nio, both of which sold at higher average prices.

The institutional root cause behind it all was the failure of "parliamentary" decision-making. On G9 version planning, He Xiaopeng had originally wanted no entry-level versions with low intelligence and short range, focusing instead on mid-to-long range; but in the end he failed to persuade the other executives, and multiple versions were launched [s6]. This exposed the fundamental flaw of the early governance model — low decision-making efficiency and unclear accountability. The company's highest decision-making body was the "general office," made up of four people: He Xiaopeng (two votes), Xia Heng (one vote), He Tao (one vote) and Brian Gu (one vote). But Xia Heng and He Tao had both come out of GAC and started the business together, so their views were almost always closely aligned. The result was a situation of "a small governing party facing a large opposition" [s6].

At a senior-executive meeting at the end of September, the atmosphere was somber as He Xiaopeng laid out the problems with the G9 launch. Lei Jun had bluntly identified XPeng's biggest problem — "You only want to be chairman; you don't want to be CEO." That sentence became the turning point in He Xiaopeng's management style [s12]. After the G9 failure, He began personally overseeing R&D, procurement and cross-functional processes; the area he managed in finest detail was procurement [s12].

After the G9 debacle, He Xiaopeng launched sweeping organizational change. He abolished the president's office and set up five virtual committees — strategy, product planning, technology planning, production-sales balance, and OTA — along with three platform-based product matrices designated E, F and H; the heads of all three platforms reported directly to He Xiaopeng, and power began to concentrate in him alone. Xia Heng stepped down as executive director at the end of 2022 [s6]. On January 30, 2023, Wang Fengying, known as the "iron lady of the car business," formally joined XPeng as president [s6].

In 2023, the personnel upheaval intensified further. That year, 85% of the heads of XPeng's thirty first-level centers were replaced, and employee turnover in the second quarter soared to 18%. At the year-end re-launch event for the G9, He Xiaopeng said candidly: "Of the twelve executives in XPeng's financial report last year, only two remain this year" [s12]. Before even joining, Wang Fengying had sharply sensed problems in XPeng's steel procurement; He Xiaopeng personally led the investigation and ultimately unearthed a supply-chain corruption ring — the procurement chief, who also held the title of president, Li Feng, was taken away by the police, and the retired manufacturing vice president Jiang Ping was also implicated. This was as much a restructuring of power as an anti-corruption campaign [s6].

In an interview with LatePost in early 2025, He Xiaopeng said: "We are still on our way out of the ICU... Today we have only won a small battle. A small win is always better than death — but it is not enough" [s12]. The sentence distills the struggle and persistence of the more than two years after the G9 failure. His office was fitted with double-glazed glass — half to shut out Wang Fengying's roaring from outside, half to keep the sound of his own outbursts from leaking out [s12]. "A chairman can afford to play the good guy, but a CEO cannot" [s12] — that line became the spiritual footnote to XPeng's organizational revolution.

### 10. 管理哲学：从'老好人'到69.3%的独裁者 / Management Philosophy: From 'the Nice Guy' to the Autocrat with 69.3%

2023年7月，大众汽车集团宣布向小鹏汽车增资约7亿美元，持有4.99%股权[s19]。这是大众在中国新能源市场节节败退后的战略选择——ID系列销量不及预期，智能化水平明显落后于中国品牌，急需一个能快速补齐智能化短板的合作伙伴。双方联合开发两款B级电动车，基于小鹏G9平台开发，预计售价20-30万元。关键在于，小鹏向大众输出CEA整车电子电气架构技术——这是中国车企首次向全球主流外资品牌输出整车核心电子架构[s18][s19]。

大众合作给小鹏带来的不仅是品牌背书，更有真金白银的收入。2025年第四季度，小鹏服务及其他收入达31.8亿元（同比增长121.9%），其中主要增量来自与大众合作的技术研发服务收入（达成重要里程碑确认收入）。这项业务的毛利率高达70.8%，直接贡献了约22.5亿元毛利——占比仅14%的非卖车收入，贡献了全公司近一半的毛利润[s7][s23]。合作已从纯电平台扩展到燃油和插混平台，技术授权收入有望持续增长[s18]。这意味着小鹏正在从“卖车”向“卖技术”转型，打开了全新的盈利空间。

与此同时，MONA品牌成为小鹏销量破局的利器。MONA源自滴滴内部的“达芬奇”造车项目，2021年至2023年滴滴投入两年时间打造，共投入上千人、资金上亿元。罗文为MONA明确了产品方向，许鹏飞的算法团队填补了智能化核心[s20]。2023年8月28日，小鹏宣布收购滴滴智能电动汽车项目资产，交易对价最高58.35亿港元，分四阶段支付，含销量对赌条款——如果每年在出行体系里卖出10万辆会有额外对价，连续两年达18万辆则滴滴持股提升至约5%[s20][s21]。

MONA最早的开发逻辑是面向网约车市场，但小鹏拿到项目后做了很大改动——掀背轿跑设计已经不适合网约车了。何小鹏在发布会上明确表示，MONA M03不会推出“的士版”，用户画像聚焦年轻人首购、智能升级换购和新生家庭增购群体[s21]。2024年8月，MONA M03以11.98万元起售价上市，上市52分钟大定破万，48小时超3万辆[s22]。到2025年7月累计交付15万台，连续11个月交付破万，稳居A级纯电轿车冠军。2025年11月累计交付突破20万辆，全年交付17.53万辆，占小鹏年度总交付量超40%[s22]。

2024年11月，小鹏P7+正式上市，进一步丰富了中高端产品线。2026年6月22日，MONA系列首款SUV车型MONA L03正式发布，定位紧凑型5座溜背轿跑SUV，提供纯电+增程双动力路线。纯电版CLTC续航505-650km，增程版纯电续航约241-257km[s22]。从MONA M03到L03，小鹏在10-15万元价格区间建立了强大的产品矩阵，成为拉动销量增长的核心引擎。

**English:** In July 2023, Volkswagen Group announced a roughly US$700 million capital injection into XPeng, taking a 4.99% equity stake [s19]. It was Volkswagen's strategic choice after one setback after another in China's new-energy market: the ID series was selling below expectations and its level of intelligentization lagged visibly behind Chinese brands, so it urgently needed a partner that could quickly close its intelligence gap. The two sides would jointly develop two B-class electric vehicles, based on XPeng's G9 platform, with expected prices of RMB 200,000-300,000. Crucially, XPeng would export its CEA vehicle electrical and electronic architecture technology to Volkswagen — the first time a Chinese carmaker had exported a vehicle's core electronic architecture to a mainstream global foreign brand [s18][s19].

The Volkswagen partnership brought XPeng not only brand endorsement but hard-cash revenue. In the fourth quarter of 2025, XPeng's services and other revenues reached RMB 3.18 billion (up 121.9% year on year), with the main increment coming from technology R&D service revenue tied to the Volkswagen cooperation (revenue recognized upon achievement of key milestones). This business carried a gross margin as high as 70.8% and directly contributed about RMB 2.25 billion in gross profit — non-vehicle-sales income, which made up only 14% of revenue, contributed nearly half of the company's total gross profit [s7][s23]. The cooperation has since expanded from pure-electric platforms to combustion and plug-in-hybrid platforms, and technology-licensing income is expected to keep growing [s18]. This means XPeng is shifting from "selling cars" to "selling technology," opening up an entirely new profit space.

Meanwhile, the MONA brand became the weapon that broke open XPeng's sales. MONA originated from Didi's in-house "Da Vinci" carmaking project, on which Didi spent two years between 2021 and 2023, deploying over a thousand people and more than RMB 100 million in funds. Luo Wen defined MONA's product direction, while Xu Pengfei's algorithm team filled in the core of its intelligence [s20]. On August 28, 2023, XPeng announced its acquisition of Didi's smart-EV project assets for consideration of up to HK$5.835 billion, payable in four phases and including sales-earn-out clauses — an additional payment if 100,000 units a year were sold through the ride-hailing system, and a rise in Didi's stake to roughly 5% if sales hit 180,000 units for two consecutive years [s20][s21].

MONA's earliest development logic had been aimed at the ride-hailing market, but after taking over the project XPeng made major changes — the hatchback-coupe design was no longer suited to taxis. At the launch event He Xiaopeng made clear that the MONA M03 would not offer a "taxi version"; its user profile focused on young first-time buyers, customers trading up for intelligence, and growing families adding a second car [s21]. In August 2024, the MONA M03 went on sale at a starting price of RMB 119,800, topping 10,000 firm orders within 52 minutes and 30,000 within 48 hours [s22]. By July 2025 cumulative deliveries reached 150,000 units, with monthly deliveries above 10,000 for eleven consecutive months, firmly holding the crown in the A-class pure-electric sedan segment. In November 2025 cumulative deliveries broke through 200,000 units; full-year deliveries came to 175,300, accounting for over 40% of XPeng's annual total [s22].

In November 2024, the XPeng P7+ was officially launched, further enriching the mid-to-high-end product line. On June 22, 2026, the MONA L03 — the first SUV in the MONA series — was officially unveiled, positioned as a compact five-seat fastback-coupe SUV offering both pure-electric and range-extended powertrains. The pure-electric version has CLTC range of 505-650 km, while the range-extended version offers roughly 241-257 km of pure-electric range [s22]. From the MONA M03 to the L03, XPeng has built a formidable product matrix in the RMB 100,000-150,000 price band, making it the core engine of sales growth.

### 11. 财务密码：43万辆交付背后的盈利真相 / The Financial Code: The Truth about Profitability behind 430,000 Deliveries

小鹏是国内最早布局智能驾驶的车企之一。2017年何小鹏全职创业时，就决心要自研自动驾驶。他亲自飞到美国，两赴硅谷邀请前特斯拉Autopilot高级工程师谷俊丽[s6]。这条自研之路走了将近十年，从XPILOT到XNGP再到VLA，每一步都代表着中国智驾技术的突破。2022年9月，城市NGP在广州行业首发，标志着中国智能驾驶进入城市场景时代[s2]。

2026年3月，小鹏图灵AI智驾（第二代VLA）正式推送，标志着从规则驱动进入大模型驱动的新阶段[s24]。第二代VLA的技术参数令人瞩目：模型参数720亿，训练数据接近1亿个视频片段，每5天完成一次全链路迭代。配合自研图灵AI芯片和定制化编译器，车端推理效率提升12倍。算力平台达2250TOPS，拥有30秒长时记忆加6秒未来场景预判能力。感知精度方面，2K纯视觉占用网络超过200万个网格重构三维空间，建模精度达5立方厘米/体素[s24]。

在与特斯拉FSD的技术路线对比中，小鹏XNGP展现出了差异化优势。两者都采用纯视觉为主的方案，但小鹏在视觉基础上增加了5个毫米波雷达辅助；特斯拉使用单一端到端大模型，而小鹏采用物理世界大模型、原生多模态方案；在记忆能力上，特斯拉是瞬时推理（十几秒），小鹏实现了30秒记忆加6秒预判。在交互风格上，小鹏采取提前语音通报加图形展示的方式，而特斯拉是沉默执行。更关键的是，小鹏深度适配中国路况，而FSD基于北美路况训练[s24]。

2026年2月，第三方媒体组织了上海到杭州到黄山再回上海的800公里环线测试，覆盖早高峰拥堵、高速匝道、山区弯道、雨雾天气等12类典型场景。结果显示：特斯拉FSD Beta 12.5总计接管15次，华为ADS 3.0总计接管2次，小鹏XNGP 5.0总计接管6次，分布相对均衡[s25]。关键场景通过率：无保护左转弯约85%，雨雾天气接管率约5%。虽然华为表现更优，但小鹏在国产品牌中处于领先水平。

何小鹏和刘先明在AMA直播中回应了“领先行业5倍”的说法——这是依据内部测试结果，同一时间、同一路段、涵盖全场景，友商平均接管次数比小鹏多4-6次。更体现为驾驶时“情绪稳定指数”提升5倍。何小鹏表示，第二代VLA的目标是打造“妈妈都喜欢的智驾”，安全第一、丝滑第二[s24]。何小鹏说：“数据是下一代汽车的新燃料，芯片和模型是新引擎。”[s26]这句话概括了他对智能驾驶未来竞争的核心判断。

**English:** XPeng was one of the first carmakers in China to lay out intelligent driving. When He Xiaopeng went into full-time entrepreneurship in 2017, he was already determined to develop autonomous driving in-house. He flew to the United States in person and traveled twice to Silicon Valley to recruit Gu Junli, a former senior engineer on Tesla's Autopilot team [s6]. That in-house path has now run for nearly a decade, from XPILOT to XNGP and on to VLA, each step representing a breakthrough in Chinese intelligent-driving technology. In September 2022, City NGP made its industry debut in Guangzhou, marking the entry of Chinese intelligent driving into the urban-scenario era [s2].

In March 2026, XPeng's Turing AI driving (second-generation VLA) was officially pushed out, marking a new stage from rule-driven to large-model-driven systems [s24]. The second-generation VLA's technical parameters are striking: 72 billion model parameters, training data approaching 100 million video clips, and one full-stack iteration completed every five days. Paired with the self-developed Turing AI chip and a customized compiler, on-vehicle inference efficiency is lifted twelvefold. The computing platform reaches 2,250 TOPS, with 30-second long-term memory plus six-second foresight of future scenes. In perception accuracy, a 2K pure-vision occupancy network reconstructs three-dimensional space from more than 2 million grids, with modeling precision down to 5 cubic centimeters per voxel [s24].

Compared with Tesla FSD on technical routes, XPeng XNGP shows differentiated advantages. Both use a vision-led approach, but XPeng adds five millimeter-wave radars as backup on top of vision; Tesla uses a single end-to-end large model, while XPeng adopts a physical-world large model with a natively multimodal scheme; on memory, Tesla reasons in the instant (a dozen or so seconds), while XPeng achieves 30-second memory plus six-second anticipation. In interaction style, XPeng uses advance voice announcements plus graphic display, whereas Tesla executes in silence. Most critically, XPeng is deeply adapted to Chinese road conditions, while FSD is trained on North American roads [s24].

In February 2026, third-party media organized an 800-kilometer loop test from Shanghai to Hangzhou to Huangshan and back to Shanghai, covering twelve categories of typical scenes including morning-peak congestion, highway ramps, mountain curves and rainy-fog weather. The results: Tesla FSD Beta 12.5 required 15 takeovers in total, Huawei ADS 3.0 required 2, and XPeng XNGP 5.0 required 6, a relatively balanced distribution [s25]. Pass rates in key scenes: roughly 85% for unprotected left turns, with a takeover rate of about 5% in rainy-fog conditions. Though Huawei performed better, XPeng stood at a leading level among domestic brands.

In an AMA livestream, He Xiaopeng and Liu Xianming responded to the claim of "five times ahead of the industry" — it was based on internal testing over the same period, on the same stretches of road and covering the full range of scenes, in which rival brands averaged 4-6 more takeovers than XPeng. Even more meaningful was a fivefold improvement in the "emotional stability index" while driving. He Xiaopeng said the goal of the second-generation VLA is to build "intelligent driving that even moms love" — safety first, smoothness second [s24]. "Data is the new fuel for the next generation of cars, and chips and models are the new engines," He said [s26]. The line sums up his core judgment on the future competition in intelligent driving.

### 12. 行业竞合：'三个苦逼'与造车新势力的江湖 / Industry Rivalry and Alliance: 'Three Wretched Sloggers' and the World of China's EV Upstarts

小鹏汇天的前身是赵德利创办的汇天科技，2013年开始研发飞行汽车。何小鹏2020年投资汇天后将其纳入小鹏体系。目前已累计投入超100亿元，团队规模超1600人（研发占比80%，硕博超40%），全球持有超1000项飞行汽车相关专利[s27]。何小鹏宣布，在量产交付前，公司高管团队将率先完成5000公里飞行测试——这种“自虐式”承诺被视为对产品安全性的极致自信[s27]。

2025年11月，小鹏科技日发布了“陆地航母”——全球首款分体式量产飞行汽车。陆行体为三轴六轮六驱增程式混合动力，CLTC续航超1000km，可停标准车位；飞行体为六轴六桨双涵道纯电垂直起降，续航约50km，最高航速200km/h。5分钟可自动完成分离或对接，C照即可驾驶陆行体，售价不超过200万元[s27][s28]。全球订单超7000台（含中东600台出口订单），计划2026年下半年启动量产交付，初期年产能5000台，规划1万台[s27]。同期发布的全倾转混电飞行汽车A868，采用全倾转旋翼设计6人座舱，预计航程500km，最高航速360km/h，面向粤港澳大湾区等城市群间30分钟跨城出行[s27]。

在机器人领域，何小鹏说：“我越来越看到高阶通用人形机器人的技术挑战跟创新是远高于智能汽车的难度，我认为至少相差20倍。”[s30]小鹏人形机器人IRON历经8年研发、7代产品，超过85%的供应链合作伙伴与小鹏汽车业务重合，成本竞争力处于领先水平，计划2026年底规模量产、2027年商用。IRON的人形设计让它可以在工厂、仓库、家庭等多种场景中工作，是小鹏物理AI战略的重要拼图[s8]。

2026年8月，小鹏机器人业务完成首轮超9亿美元融资，投后估值约430亿元人民币（约63亿美元）。交易完成后，小鹏汽车仍持有73.8%股权，IDG资本持股4.72%，阿里、腾讯、高榕创投分别持股1.57%[s8][s10]。这标志着机器人业务从内部孵化走向独立发展。机器人业务亏损情况：2024年净亏损8700万元，2025年净亏损3.69亿元[s8]。

在何小鹏眼中，汽车只是“四个轮子的机器人”，是漫长技术演进中的一个节点而非终点。汽车、飞行汽车、机器人共享物理AI技术底座，构成小鹏的“三条曲线”战略[s9]。他提出用10年时间将小鹏打造为全球物理AI公司，海外收入占比达到一半。“硬软结合，硬在前”[s9]——这是何小鹏对未来科技公司的判断。2026年6月，他在担任小鹏集团CEO的同时兼任机器人业务CEO，把花了几年的旧自动驾驶体系全部砍掉，押注VLA大模型路线——当时每个月要投入3个亿[s8]。这些决定都是他一个人拍板的——69.3%的投票权给了他这种魄力，也给了他这种责任。

**English:** XPeng AeroHT grew out of Huitian Technology, founded by Zhao Deli, which began developing flying cars in 2013. After He Xiaopeng invested in Huitian in 2020, he brought it into the XPeng fold. Cumulative investment has now exceeded RMB 10 billion, the team numbers more than 1,600 people (80% in R&D, over 40% holding master's or doctoral degrees), and the company holds more than 1,000 flying-car-related patents worldwide [s27]. He Xiaopeng has announced that, before volume deliveries begin, the company's executive team will first complete a 5,000-kilometer flight test themselves — a "self-flagellating" commitment taken as ultimate confidence in the product's safety [s27].

At XPeng's Tech Day in November 2025, the company unveiled the "Land Aircraft Carrier" — the world's first split-configuration mass-produced flying car. The ground module is a three-axle, six-wheel, six-drive range-extended hybrid with CLTC range of over 1,000 km that can park in a standard parking space; the flight module is a six-axle, six-rotor, dual-ducted all-electric VTOL craft with range of about 50 km and a top speed of 200 km/h. Separation or docking can be completed automatically in five minutes; the ground module can be driven on an ordinary Class C driver's license, and the price is capped at RMB 2 million [s27][s28]. Global orders exceed 7,000 units (including 600 export orders from the Middle East); volume deliveries are planned to begin in the second half of 2026, with initial annual capacity of 5,000 units and a planned 10,000 [s27]. Unveiled at the same time was the A868 full-tilt-rotor hybrid-electric flying car, with a full-tilt-rotor design and a six-seat cabin, expected range of 500 km and top speed of 360 km/h, aimed at 30-minute intercity travel across city clusters such as the Greater Bay Area [s27].

In robotics, He Xiaopeng says: "I increasingly see that the technical challenges and innovation involved in high-end general-purpose humanoid robots are far harder than for smart cars — at least twenty times harder, in my view" [s30]. XPeng's humanoid robot IRON has been through eight years of R&D and seven product generations; over 85% of its supply-chain partners overlap with XPeng's auto business, giving it leading cost competitiveness, with mass production planned for the end of 2026 and commercial deployment in 2027. IRON's humanoid form lets it work in factories, warehouses, homes and many other settings, making it a key piece of XPeng's physical-AI strategy [s8].

In August 2026, XPeng's robotics business closed a first funding round of over US$900 million, at a post-money valuation of about RMB 43 billion (roughly US$6.3 billion). After the transaction XPeng Inc. still holds a 73.8% stake, IDG Capital holds 4.72%, and Alibaba, Tencent and Gaorong Capital each hold 1.57% [s8][s10]. It marked the robotics business's move from internal incubation to independent development. The robotics business's losses: a net loss of RMB 87 million in 2024 and RMB 369 million in 2025 [s8].

In He Xiaopeng's eyes, the car is merely "a robot on four wheels" — a node in a long arc of technological evolution rather than the end point. Cars, flying cars and robots share the physical-AI technology base, forming XPeng's "three curves" strategy [s9]. He has set a goal of building XPeng into a global physical-AI company within ten years, with overseas revenue reaching half of the total. "Hardware and software combined, with hardware first" [s9] — that is his judgment on the technology company of the future. In June 2026, while remaining CEO of XPeng Group, he also took on the role of CEO of the robotics business; he scrapped the old autonomous-driving system that had taken years to build and bet on the VLA large-model route — at the time consuming RMB 300 million a month [s8]. These decisions were all made by him alone: his 69.3% voting stake gave him both the nerve for such calls and the responsibility that comes with them.

## 卷尾 Editorial Conclusion

何小鹏的故事本质上是一个关于“连续创业”的寓言。第一次创业，他用UC浏览器证明了程序员也能做出四亿用户的产品，最终43亿美元卖身阿里创下中国互联网最大并购纪录——那笔交易让UC团队获得了1230万股阿里限制性股票，占阿里总股本0.5%[s5]。

第二次创业，他用小鹏汽车证明了一个互联网人可以在制造业的泥潭里爬起来。即便G9折戟、高管出走、12位核心高管只剩2位、销量腰斩——他依然能从ICU里走出来[s12]。王凤英这把“手术刀”帮他铲平了山头，供应链反腐帮他清理了蛀虫，大众7亿美元的技术订单帮他找到了新的盈利模式[s6][s19]。

如今，他正试图完成第三次跨越——从地面到天空、从轮子到双腿、从汽车公司到物理AI平台。飞行汽车“陆地航母”全球订单超7000台，机器人IRON估值430亿元完成超9亿美元首轮融资，图灵AI智驾720亿参数大模型开始领跑行业[s27][s10][s24]。490亿元的财富没有让他停下脚步，69.3%的投票权给了他“一个人说了算”的底气[s18]。

但真正的考验不是财富和控制权，而是他能否在汽车、飞行汽车、机器人三条战线上同时兑现承诺。飞行汽车累计投入超100亿，机器人还在亏损（2025年亏3.69亿），2026上半年小鹏净亏损扩大至31亿元[s8]。正如他自己所说：“最大的坎儿是不知道还有多少个坎儿。”[s29]这位归来的少年，仍在路上。而在他的蓝图里，汽车只是“四个轮子的机器人”——是漫长技术演进中的一个节点，而非终点[s9]。

He Xiaopeng's story is, at bottom, a parable about serial entrepreneurship. In his first venture, UC Browser proved that a programmer could build a product used by 400 million people, and its sale to Alibaba for US$4.35 billion set a record as the largest acquisition in Chinese internet history — a deal that handed the UC team 12.3 million restricted Alibaba shares, equal to 0.5% of Alibaba's total share capital [s5].

In his second venture, XPeng showed that an internet man could claw his way up out of the mire of manufacturing. Even after the G9 debacle, the exodus of executives — only two of twelve core executives remained — and a halving of sales, he still managed to walk out of the ICU [s12]. Wang Fengying served as the "scalpel" that flattened internal fiefdoms; the supply-chain anti-corruption campaign cleared out the parasites; and Volkswagen's US$700 million technology order opened up a new profit model [s6][s19].

Now he is attempting a third leap — from the ground to the sky, from wheels to legs, from a car company to a physical-AI platform. The "Land Aircraft Carrier" flying car has drawn more than 7,000 orders worldwide; the IRON robot business closed a first funding round of over US$900 million at a valuation of RMB 43 billion; and the Turing AI driving model, with its 72 billion parameters, has begun to lead the industry [s27][s10][s24]. A fortune of RMB 49 billion has not slowed him down, and his 69.3% voting stake underwrites his ability to "call the shots alone" [s18].

Yet the real test is not wealth or control, but whether he can deliver simultaneously on three fronts — cars, flying cars and robots. The flying-car program has absorbed more than RMB 10 billion in cumulative investment; the robotics business is still loss-making (RMB 369 million in 2025); and XPeng's net loss widened to RMB 3.1 billion in the first half of 2026 [s8]. As he himself puts it: "The biggest hurdle is not knowing how many more hurdles lie ahead" [s29]. The boy who came back is still on the road. And in his blueprint, the car is merely "a robot on four wheels" — one node in a long arc of technological evolution, not the finish line [s9].

## 金句 Pull Quote

> 现在是从ICU往外走的路上……我们今天只是小赢，小赢总比死好，但也不够。

> "We are still on our way out of the ICU... Today we have only won a small battle. A small win is always better than death — but it is not enough."
— He Xiaopeng, in an interview with LatePost, February 2025

## 履历时间线 Career Timeline

- **1977** 何小鹏出生于湖北省黄石市一个普通工人… / Born in Huangshi, Hubei
  - 何小鹏出生于湖北省黄石市一个普通工人家庭，祖籍洪湖市府场镇。父母都是普通职员，家境平凡。这座长江中游的工业城市给了他最初的理工科启蒙——父亲的工厂里的机器轰鸣，构成了一个少年对"技术"最早的感知。[s1]
  - EN: He Xiaopeng was born into an ordinary working-class family in Huangshi, Hubei province, with his ancestral home in Fuchang town, Honghu. Both parents were ordinary staff and the family circumstances were modest. This industrial city on the middle reaches of the Yangtze gave him his earliest exposure to science and engineering — the roar of machinery in his father's factory formed a boy's first perception of "technology." [s1]
- **1995** 考入华南理工大学计算机专业。从黄石到… / Enters South China University of Technology
  - 考入华南理工大学计算机专业。从黄石到广州，是他第一次离开家乡。广州的开放氛围和商业活力，深深影响了这个内向的湖北少年。华南理工的四年成为他人生中最重要的阶段之一。[s2]
  - EN: He was admitted to South China University of Technology to study computer science. Moving from Huangshi to Guangzhou was the first time he had left his hometown. The open atmosphere and commercial vitality of Guangzhou deeply influenced the introverted Hubei teenager. His four years at SCUT became one of the most important phases of his life. [s2]
- **1997** 暑假带领同学将西18宿舍楼全部连接上… / Wires whole dormitory to the education network
  - 暑假带领同学将西18宿舍楼全部连接上教育网，这可能是中国大学最早整栋通网的宿舍之一。当时大多数学生还不知道"上网"是什么概念，他已经在实践中摸索网络架构了，展现了对技术的敏锐度和行动力。[s1]
  - EN: During the summer break he led his classmates in connecting every computer in the West Building 18 dormitory to the education and research network — possibly one of the earliest university dormitories in China to have an entire building online. At a time when most students had no idea what "going online" even meant, he was already experimenting with network architecture in practice, displaying a sharp feel for technology and a bias for action. [s1]
- **1999** 毕业于华南理工大学计算机科学专业，获… / Graduates; joins AsiaInfo in Guangzhou
  - 毕业于华南理工大学计算机科学专业，获学士学位。同年加入亚信科技广州分公司，这是中国最早的互联网基础设施公司之一。他在这里度过了五年半时光，从程序员成长为项目管理者。[s2]
  - EN: He graduated from South China University of Technology with a bachelor's degree in computer science. The same year he joined the Guangzhou branch of AsiaInfo Technologies, one of China's earliest internet-infrastructure companies. He spent five and a half years there, growing from a programmer into a project manager. [s2]
- **2000** 亚信集团在纳斯达克上市，何小鹏获15… / AsiaInfo lists on Nasdaq
  - 亚信集团在纳斯达克上市，何小鹏获1500股原始股，市值约3万美元。这是他第一次接触资本市场，也在心里埋下了创业的种子。[s1]
  - EN: When AsiaInfo Group listed on Nasdaq, He Xiaopeng received 1,500 founder shares worth about US$30,000. It was his first contact with capital markets, and it planted the seed of entrepreneurship in his mind. [s1]
- **2004** 27岁从亚信离职，与校友梁捷共同创办… / Founds UCWeb with Liang Jie
  - 27岁从亚信离职，与校友梁捷共同创办UC优视。最初做移动邮件UCMail，后转向手机浏览器UCWeb。网易丁磊是第一个贵人——直接提供80万借款、网易服务器和办公室。UCWeb最初是UCMail的副产品，却成了改变公司命运的核心产品。[s3]
  - EN: At 27 he left AsiaInfo and co-founded UCWeb with fellow alumnus Liang Jie. They began with mobile email, UCMail, before pivoting to the UCWeb mobile browser. NetEase's Ding Lei was his first benefactor — directly providing an RMB 800,000 loan, NetEase servers and an office. UCWeb itself began as a by-product of UCMail, yet became the core product that changed the company's fate. [s3]
- **2006** UC资金链濒临断裂，创业第一年搬了6… / Cash crisis; Lei Jun invests RMB 4m
  - UC资金链濒临断裂，创业第一年搬了6次家，每月四处借钱给员工发工资。红杉资本考察时，两人竟说没钱买机票去北京。经李学凌牵线认识俞永福，雷军以"俞永福去我就投"为条件天使投资400万元。[s3][s11]
  - EN: UC's cash chain came to the brink of collapse; in the start-up's first year the company moved offices six times, and every month He borrowed money from all quarters to make payroll. When Sequoia Capital visited, the two founders actually said they could not afford plane tickets to Beijing. Through an introduction by Li Xueling he met Yu Yongfu, and Lei Jun made a RMB 4 million angel investment on condition that "if Yu Yongfu goes, I'll invest." [s3][s11]
- **2007** 晨兴投资和联创策源联合投资1000万… / US$10m Series A; the troika takes shape
  - 晨兴投资和联创策源联合投资1000万美元A轮。俞永福推动三个关键决策：聚焦手机浏览器、拒绝企业业务只做To C、融资门槛定为1000万美元。"三驾马车"格局确立——何小鹏管产品创意、梁捷管技术、俞永福管管理资本。[s1][s4]
  - EN: Morningside Ventures and GGV Capital co-led a US$10 million Series A round. Yu Yongfu drove three key decisions: focus on the mobile browser, refuse enterprise business and serve only consumers, and set the financing bar at US$10 million. The "troika" structure was established — He Xiaopeng over product ideas, Liang Jie over technology, and Yu Yongfu over management and capital. [s1][s4]
- **2012** UC浏览器用户突破4亿，成为BAT追… / UC Browser tops 400 million users
  - UC浏览器用户突破4亿，成为BAT追逐的对象。百度开价20亿美元收购被拒，俞永福称"UC是非卖品"。马化腾也曾示好，但条件谈不拢。最终选择阿里，因为马云给了UC最大的自主空间。[s5]
  - EN: UC Browser's user base broke through 400 million, making it the object of pursuit by the BAT trio (Baidu, Alibaba, Tencent). Baidu offered US$2 billion to acquire it and was turned down, with Yu Yongfu declaring "UC is not for sale." Pony Ma of Tencent also courted the company but terms could not be agreed. UC ultimately chose Alibaba, because Jack Ma offered the greatest operational independence. [s5]
- **2014** 阿里巴巴以约43.5亿美元全资收购U… / Invests in XPeng predecessor Orange Xing
  - 阿里巴巴以约43.5亿美元全资收购UC，创当时中国互联网最大并购。交易以股票为主现金为辅，UC团队获1230万股阿里限制性股票。俞永福进入阿里8人战略决策委员会。何小鹏出任移动事业群总裁、阿里游戏董事长、土豆总裁。[s5]
  - EN: After Tesla opened its patents, he proposed building cars inside Alibaba without success. The same year, together with internet luminaries including Li Xueling, Fu Sheng, Wu Xiaoguang and Yu Yongfu, he invested in "Orange Xing Zhidong," founded by Xia Heng and He Tao. The company wanted to call itself "Orange Auto" but could not secure the trademark, and was ultimately named after He Xiaopeng. He personally put up about RMB 300 million in start-up capital. [s6]
- **2014** 特斯拉开放专利后在阿里内部建议造车未… / Invests in XPeng predecessor Orange Xing
  - 特斯拉开放专利后在阿里内部建议造车未果。同年联合李学凌、傅盛、吴宵光、俞永福等互联网大佬，投资夏珩、何涛创立的"橙行智动"。本想叫"橙子汽车"但商标申请不下来，最终用何小鹏名字命名。自掏约3亿元启动资金。[s6]
  - EN: After Tesla opened its patents, he proposed building cars inside Alibaba without success. The same year, together with internet luminaries including Li Xueling, Fu Sheng, Wu Xiaoguang and Yu Yongfu, he invested in "Orange Xing Zhidong," founded by Xia Heng and He Tao. The company wanted to call itself "Orange Auto" but could not secure the trademark, and was ultimately named after He Xiaopeng. He personally put up about RMB 300 million in start-up capital. [s6]
- **2017** UC十三岁生日当天正式从阿里离职。俞… / Joins XPeng as chairman
  - UC十三岁生日当天正式从阿里离职。俞永福称其将"荣退"并"开启追梦之旅"。朋友圈发文："创业一轮回，苦辣酸甜咸，归来仍是少年。"没有带走阿里部门一个人——"要对老东家讲义气。"[s1]
  - EN: He formally joined XPeng as chairman. The company already had a headcount of 300. His first major decision was to tear up and rebuild the G3 Beta version — judging that the existing version fell short of mass-production standard, a call that delayed launch by nearly a year. He styled himself the "chief product experience officer." [s2][s6][s17]
- **2017** 正式加入小鹏汽车出任董事长。当时公司… / Joins XPeng as chairman
  - 正式加入小鹏汽车出任董事长。当时公司已有300人规模。他的第一个重大决定是推倒重做G3 Beta版——认为现有版本达不到量产标准，为此上市推迟近一年。自称"首席产品体验官"。[s2][s6][s17]
  - EN: He formally joined XPeng as chairman. The company already had a headcount of 300. His first major decision was to tear up and rebuild the G3 Beta version — judging that the existing version fell short of mass-production standard, a call that delayed launch by nearly a year. He styled himself the "chief product experience officer." [s2][s6][s17]
- **2018** 首款量产车小鹏G3正式上市发布并启动… / First production car, the G3, launched
  - 首款量产车小鹏G3正式上市发布并启动交付，定位紧凑型纯电SUV。从广州天河区商住两用房的创业起步到第一款量产车下线，小鹏用了不到四年。[s2]
  - EN: XPeng's first production car, the G3, was officially launched with deliveries beginning, positioned as a compact all-electric SUV. From its beginnings in a combined residential-commercial apartment in Tianhe, Guangzhou, to the first production car rolling off the line, XPeng took less than four years. [s2]
- **2020** 旗舰纯电轿车P7上市，设计语言和智能… / Lists on the NYSE as XPEV
  - 旗舰纯电轿车P7上市，设计语言和智能化配置获市场广泛认可，成为小鹏早期最成功的车型之一，被视为品牌向上突破的关键产品。[s15]
  - EN: XPeng listed on the New York Stock Exchange under the ticker XPEV at an offer price of US$15 per ADS, corresponding to a market value of US$10.82 billion — the largest new-energy-vehicle IPO in the world at the time. From a residential-commercial apartment to the New York Stock Exchange, XPeng took less than six years. [s15]
- **2020** 小鹏汽车在纽交所上市（XPEV），发… / Lists on the NYSE as XPEV
  - 小鹏汽车在纽交所上市（XPEV），发行价15美元/ADS，对应市值108.2亿美元，为当时全球新能源汽车最大IPO。从商住两用房到纽约证券交易所，小鹏用了不到六年。[s15]
  - EN: XPeng listed on the New York Stock Exchange under the ticker XPEV at an offer price of US$15 per ADS, corresponding to a market value of US$10.82 billion — the largest new-energy-vehicle IPO in the world at the time. From a residential-commercial apartment to the New York Stock Exchange, XPeng took less than six years. [s15]
- **2021** 在香港联交所挂牌上市，成为首个在纽约… / P5: world's first mass-produced lidar car
  - 在香港联交所挂牌上市，成为首个在纽约和香港两地双重主要上市的中国造车新势力，拓宽了融资渠道和国际影响力。[s2]
  - EN: The XPeng P5 was officially launched as the world's first mass-produced smart car equipped with lidar, marking a leading position in intelligent-driving hardware. It was also an important milestone in XPeng's strategy of developing intelligent driving in-house. [s2]
- **2021** 小鹏P5正式上市，成为全球首款量产激… / P5: world's first mass-produced lidar car
  - 小鹏P5正式上市，成为全球首款量产激光雷达智能汽车，标志着在智能驾驶硬件领域的领先地位。这也是小鹏自研智驾战略的重要里程碑。[s2]
  - EN: The XPeng P5 was officially launched as the world's first mass-produced smart car equipped with lidar, marking a leading position in intelligent-driving hardware. It was also an important milestone in XPeng's strategy of developing intelligent driving in-house. [s2]
- **2022** G9首次上市因SKU复杂、定价混乱导… / G9 launch fiasco
  - G9首次上市因SKU复杂、定价混乱导致惨败——48小时内被迫改价。原定三个月3.5万辆目标，实际季度仅交付6189辆。整个小鹏销量被带崩，从上半年近7万台滑到下半年两万多。暴露了"议会制"决策模式的根本缺陷。[s6]
  - EN: The G9's first launch ended in disaster amid complex SKUs and chaotic pricing — the company was forced into repricing within 48 hours. Against an original target of 35,000 units in three months, actual quarterly deliveries were only 6,189. XPeng's overall sales were dragged down, sliding from nearly 70,000 units in the first half to just over 20,000 in the second. The episode exposed the fundamental flaw of the "parliamentary" decision-making model. [s6]
- **2023** "车圈铁娘子"王凤英正式加入出任总裁… / Acquires Didi Da Vinci EV project
  - "车圈铁娘子"王凤英正式加入出任总裁。她在长城汽车31年，从销售成长为总经理。加入后将铁腕手段一并带来，开始整顿产品规划、销售体系和人事架构。[s12]
  - EN: XPeng reached a strategic cooperation with Didi, acquiring the assets of its smart-EV "Da Vinci" project to build the MONA brand. Consideration ran as high as HK$5.835 billion, payable in four phases, with sales earn-out clauses tied to annual volumes of 100,000 and 180,000 vehicles. [s20][s21]
- **2023** 大众汽车集团向小鹏增资约7亿美元持股… / Acquires Didi Da Vinci EV project
  - 大众汽车集团向小鹏增资约7亿美元持股4.99%，双方联合开发B级电动车。小鹏向大众输出CEA整车电子电气架构——中国车企首次向全球主流外资品牌输出核心架构技术。[s19]
  - EN: XPeng reached a strategic cooperation with Didi, acquiring the assets of its smart-EV "Da Vinci" project to build the MONA brand. Consideration ran as high as HK$5.835 billion, payable in four phases, with sales earn-out clauses tied to annual volumes of 100,000 and 180,000 vehicles. [s20][s21]
- **2023** 与滴滴达成战略合作，收购其智能电动汽… / Acquires Didi Da Vinci EV project
  - 与滴滴达成战略合作，收购其智能电动汽车"达芬奇"项目资产，打造MONA品牌。对价最高58.35亿港元分四阶段支付，含年销10万/18万辆的销量对赌条款。[s20][s21]
  - EN: XPeng reached a strategic cooperation with Didi, acquiring the assets of its smart-EV "Da Vinci" project to build the MONA brand. Consideration ran as high as HK$5.835 billion, payable in four phases, with sales earn-out clauses tied to annual volumes of 100,000 and 180,000 vehicles. [s20][s21]
- **2024** MONA M03以11.98万元起售… / MONA M03 launch — an instant hit
  - MONA M03以11.98万元起售价上市，52分钟大定破万，48小时超3万辆。连续11个月交付破万，稳居A级纯电轿车冠军。从网约车项目到C端爆款，完成了一次漂亮的转型。[s22]
  - EN: The MONA M03 launched at a starting price of RMB 119,800, topping 10,000 firm orders in 52 minutes and 30,000 in 48 hours. It delivered more than 10,000 units a month for eleven consecutive months, firmly holding the A-class pure-electric sedan crown. From a ride-hailing project to a consumer hit, it completed a beautiful transformation. [s22]
- **2025** 第四季度首次实现单季度盈利3.8亿元… / A868 tilt-rotor unveiled; Land Aircraft Carrier trial run
  - 第四季度首次实现单季度盈利3.8亿元，毛利率21.3%创历史新高。但盈利高度依赖技术授权收入（服务业务毛利率70.8%贡献近半毛利）和约8.4亿元政府补贴，核心经营利润约-9.2亿元。[s7][s23]
  - EN: XPeng AeroHT unveiled the A868 full-tilt-rotor hybrid-electric flying car (500 km range, 360 km/h top speed), and the first flight module of the Land Aircraft Carrier came off the trial-production line at its mass-production plant. Cumulative investment exceeds RMB 10 billion, the team numbers 1,600, and global orders top 7,000 units. [s27]
- **2025** 全年总交付429,445台同比增长1… / A868 tilt-rotor unveiled; Land Aircraft Carrier trial run
  - 全年总交付429,445台同比增长125.9%，海外交付45,008台同比增长96%。11月第100万台整车下线，MONA M03累计交付超20万辆。胡润百富榜第114位，财富480亿元，造车新势力首富。[s7]
  - EN: XPeng AeroHT unveiled the A868 full-tilt-rotor hybrid-electric flying car (500 km range, 360 km/h top speed), and the first flight module of the Land Aircraft Carrier came off the trial-production line at its mass-production plant. Cumulative investment exceeds RMB 10 billion, the team numbers 1,600, and global orders top 7,000 units. [s27]
- **2025** 小鹏汇天发布全倾转混电飞行汽车A86… / A868 tilt-rotor unveiled; Land Aircraft Carrier trial run
  - 小鹏汇天发布全倾转混电飞行汽车A868（航程500km时速360km/h），陆地航母量产工厂试产下线首台飞行器。累计投入超100亿，团队1600人，全球订单超7000台。[s27]
  - EN: XPeng AeroHT unveiled the A868 full-tilt-rotor hybrid-electric flying car (500 km range, 360 km/h top speed), and the first flight module of the Land Aircraft Carrier came off the trial-production line at its mass-production plant. Cumulative investment exceeds RMB 10 billion, the team numbers 1,600, and global orders top 7,000 units. [s27]
- **2026** 第二代VLA大模型正式推送，720亿… / Robotics unit raises over US$900m
  - 第二代VLA大模型正式推送，720亿参数、1亿视频训练片段、每5天全链路迭代。配合自研图灵芯片车端推理效率提升12倍。800公里实测接管6次，优于特斯拉FSD的15次。目标打造"妈妈都喜欢的智驾"。[s24][s25]
  - EN: The robotics business closed a first funding round of over US$900 million at a post-money valuation of about RMB 43 billion. XPeng still holds 73.8%, IDG Capital 4.72%, and Alibaba and Tencent 1.57% each. The IRON humanoid robot is planned for mass production by the end of 2026 and commercial deployment in 2027. [s10]
- **2026** 获聘蚂蚁集团独立董事，同时兼任机器人… / Robotics unit raises over US$900m
  - 获聘蚂蚁集团独立董事，同时兼任机器人业务CEO。砍掉花"小几十个亿"的旧自动驾驶体系押注VLA大模型，推动从汽车公司向物理AI公司转型。69.3%投票权让他实现了"一个人说了算"。[s8]
  - EN: The robotics business closed a first funding round of over US$900 million at a post-money valuation of about RMB 43 billion. XPeng still holds 73.8%, IDG Capital 4.72%, and Alibaba and Tencent 1.57% each. The IRON humanoid robot is planned for mass production by the end of 2026 and commercial deployment in 2027. [s10]
- **2026** 机器人业务完成首轮超9亿美元融资，投… / Robotics unit raises over US$900m
  - 机器人业务完成首轮超9亿美元融资，投后估值约430亿元。小鹏仍持股73.8%，IDG资本4.72%，阿里腾讯各1.57%。IRON人形机器人计划2026年底规模量产、2027年商用。[s10]
  - EN: The robotics business closed a first funding round of over US$900 million at a post-money valuation of about RMB 43 billion. XPeng still holds 73.8%, IDG Capital 4.72%, and Alibaba and Tencent 1.57% each. The IRON humanoid robot is planned for mass production by the end of 2026 and commercial deployment in 2027. [s10]

## 常问问答 FAQ

**Q1: 何小鹏为什么从阿里离职去造车？**

A: 何小鹏从阿里离职造车是多重因素叠加的结果。第一，在阿里高管年会上，自己做了十年的UC'只出现了一次'，强烈的失落感让他感受到'中年危机'[s7]。第二，2014年马斯克开放特斯拉专利后他在阿里内部建议造车未果[s7]。第三，GGV符绩勋的催促——'窗口期不容错过'[s1]。第四，儿子出生时的触动——想做些事情让儿子将来问起时能骄傲回答[s7]。第五，连续创业者的内心驱动——'不折腾自己，人生太没有意义了'[s7]。2017年8月22日UC十三岁生日当天他正式离职，朋友圈发文'归来仍是少年'。这段经历后来成为他'连续创业者'身份的核心注脚。

**Q1 (EN): Why did He Xiaopeng leave Alibaba to build cars?**

A: He Xiaopeng's departure from Alibaba to make cars was the product of several factors converging at once. First, at Alibaba's annual senior-executive conference, the name of UC — the company he had built for ten years — was "mentioned only once," leaving him with a sharp sense of letdown that he described as a "mid-life crisis" [s7]. Second, after Elon Musk opened Tesla's patents in 2014, he proposed building cars inside Alibaba but failed to win support [s7]. Third came the urging of GGV's Hany Tan (Fu Jixun) — "the window of opportunity must not be missed" [s1]. Fourth was the emotional jolt of his son's birth: he wanted to do something that, when his son asked about it years later, he could answer with pride [s7]. Fifth was the inner drive of a serial entrepreneur — "if you don't keep pushing yourself, life is simply too meaningless" [s7]. On August 22, 2017, UC's thirteenth birthday, he formally resigned, posting on WeChat Moments that "the boy who comes back is still a boy." The episode later became the central footnote to his identity as a serial entrepreneur.

**Q2: 何小鹏和雷军是什么关系？**

A: 雷军是何小鹏的天使投资人兼导师，两人均为湖北人。雷军2006年底以天使身份给UC投了400万元，条件是'俞永福去我就投'[s3]。2007年雷军出任UC董事长。小米集团也参与了小鹏汽车C轮融资。雷军曾直言小鹏最大的问题是'你只想做董事长，不想做CEO'，这句话成为何小鹏管理风格转变的关键转折点[s8]。两人亦师亦友，关系持续二十余年。

**Q2 (EN): What is the relationship between He Xiaopeng and Lei Jun?**

A: Lei Jun is He Xiaopeng's angel investor and mentor, and both men hail from Hubei province. At the end of 2006, Lei invested RMB 4 million in UC as an angel, on the single condition that "if Yu Yongfu goes, I'll invest" [s3]. In 2007 Lei became UC's chairman. Xiaomi Group also took part in XPeng's Series C financing. Lei Jun once told He bluntly that his biggest problem was that "you only want to be chairman, not CEO" — a remark that became the key turning point in He Xiaopeng's transformation as a manager [s8]. The two have been both teacher and friend for more than twenty years.

**Q3: 小鹏汽车名字的由来是什么？**

A: 小鹏汽车最初叫'橙行智动'，由夏珩、何涛、杨春雷三位广汽出身的工程师创立。本想取名'橙子汽车'但没拿下商标。所有高管的名字都被拿去申请了一遍商标，都申请不下来，最后轮到何小鹏的名字才搞定，于是有了'小鹏汽车'这个名字[s8]。何小鹏最初只是天使投资人，2017年才全职加入出任董事长。

**Q3 (EN): How did XPeng get its name?**

A: XPeng was originally called "Orange Xing Zhidong," founded by three engineers with backgrounds at GAC — Xia Heng, He Tao and Yang Chunlei. The founders wanted to name it "Orange Auto" but could not secure the trademark. They filed trademark applications using every executive's name in turn, and every single one was rejected; only when it finally came to He Xiaopeng's name did the application go through — and so the company became "XPeng" [s8]. He Xiaopeng himself was initially only an angel investor and did not join full-time as chairman until 2017.

**Q4: 何小鹏持有小鹏汽车多少股份？投票权有多大？**

A: 截至2026年3月31日，何小鹏持有小鹏集团18.8%的股权，但拥有69.3%的投票权。这得益于同股不同权设计：A类股1股1票，B类股1股10票。对比：理想李想持股21.7%、投票权68.7%；蔚来李斌持股7.1%、投票权34%[s8]。何小鹏的投票权在新势力三杰中最高，确保了他对公司战略方向的绝对控制权。这种同股不同权架构确保了创始人在融资稀释后仍能掌控公司方向。

**Q4 (EN): How much equity does He Xiaopeng hold in XPeng, and how large is his voting power?**

A: As of March 31, 2026, He Xiaopeng holds 18.8% of XPeng Group's equity but commands 69.3% of its voting rights. This rests on a dual-class share structure: Class A shares carry one vote per share, while Class B shares carry ten. For comparison, Li Auto's Li Xiang holds 21.7% of equity with 68.7% of voting rights, and Nio's Li Bin holds 7.1% with 34% of voting rights [s8]. He Xiaopeng's voting power is the highest among the three EV-upstart leaders, guaranteeing him absolute control over the company's strategic direction. The dual-class structure ensures that a founder can still steer the company after repeated rounds of financing dilution.

**Q5: 小鹏汽车和大众是什么合作关系？**

A: 大众汽车2023年7月以约7亿美元收购小鹏4.99%股权，双方联合开发两款B级电动车。更重要的是，小鹏向大众输出CEA电子电气架构技术，合作已从纯电扩展到燃油和插混平台[s9]。技术授权收入成为重要利润来源——2025Q4服务及其他业务毛利率70.8%，贡献了全公司近一半毛利润[s10]。

**Q5 (EN): What is the partnership between XPeng and Volkswagen?**

A: In July 2023, Volkswagen Group acquired a 4.99% stake in XPeng for roughly US$700 million, and the two sides are jointly developing two B-class electric vehicles. More importantly, XPeng exports its CEA electrical and electronic architecture technology to Volkswagen, and the cooperation has expanded from pure-electric platforms to combustion and plug-in-hybrid platforms [s9]. Technology-licensing income has become a major profit source: in Q4 2025 the services and other business carried a 70.8% gross margin and contributed nearly half of the company's total gross profit [s10].

**Q6: 小鹏汽车何时实现盈利？盈利质量如何？**

A: 小鹏汽车2025年Q4首次实现单季度盈利3.8亿元，毛利率21.3%。但盈利高度依赖高毛利技术授权收入（服务业务毛利率70.8%）、碳积分收益和约8.4亿元政府补贴。核心经营利润约-9.2亿元，汽车主业造血能力仍在提升但尚未转正。全年净亏损11.4亿元同比收窄80.3%。2026上半年净亏损扩大至约31亿元[s8][s10]。

**Q6 (EN): When did XPeng become profitable, and what is the quality of its earnings?**

A: XPeng posted its first-ever quarterly profit of RMB 380 million in Q4 2025, with a gross margin of 21.3%. The profit, however, leaned heavily on high-margin technology-licensing income (the services business ran at a 70.8% gross margin), carbon-credit gains and roughly RMB 840 million in government subsidies. Core operating profit was around minus RMB 920 million: the core vehicle business's ability to generate cash was still improving but had not yet turned positive. For the full year the net loss was RMB 1.14 billion, narrowing by 80.3% year on year; in the first half of 2026 the net loss widened to about RMB 3.1 billion [s8][s10].

**Q7: MONA品牌和小鹏是什么关系？**

A: MONA源自滴滴内部'达芬奇'造车项目。2023年8月小鹏以最高58.35亿港元股份对价收购，分四阶段支付含销量对赌条款。MONA主打10-15万元价格区间年轻用户市场，首款M03于2024年8月上市起售价11.98万元，截至2025年11月累计交付超20万辆。2025年全年MONA交付17.53万辆占总交付超40%[s9][s10]。

**Q7 (EN): What is the relationship between the MONA brand and XPeng?**

A: MONA originated from Didi's in-house "Da Vinci" carmaking project. In August 2023, XPeng acquired it for share consideration of up to HK$5.835 billion, payable in four phases and including sales earn-out clauses. MONA targets young users in the RMB 100,000-150,000 price band; its first model, the M03, launched in August 2024 at a starting price of RMB 119,800, with cumulative deliveries exceeding 200,000 units by November 2025. For full-year 2025, MONA delivered 175,300 vehicles, accounting for over 40% of XPeng's total deliveries [s9][s10].

**Q8: G9上市失败的原因是什么？**

A: G9 2022年9月首次上市失败有三大原因：一是SKU配置极其复杂，消费者选车像做数学题；二是选配逻辑混乱，很多功能需额外加钱；三是最终价格高于市场预期。48小时内被迫紧急改价，季度仅交付6189辆（原定3.5万辆）。深层原因是早期'议会制'决策模式效率低下。G9失利成为何小鹏管理风格从'无为而治'转向'铁腕'的转折点[s8]。

**Q8 (EN): What caused the failed launch of the G9?**

A: The G9's first launch, in September 2022, failed for three main reasons: first, its SKU and configuration scheme was so complex that choosing a car felt like solving a math problem; second, the options logic was chaotic, with many features requiring extra payment; and third, the final prices came in above market expectations. The company was forced into emergency repricing within 48 hours, and quarterly deliveries reached only 6,189 units against an original target of 35,000. The deeper cause was the inefficiency of the early "parliamentary" decision-making model. The G9 setback became the turning point at which He Xiaopeng's management style shifted from laissez-faire to iron-fisted rule [s8].

**Q9: 何小鹏为什么要做飞行汽车和机器人？**

A: 何小鹏认为未来有两类巨型科技公司：全数字化公司和硬软结合的全球化公司。汽车、飞行汽车、机器人共享物理AI技术底座，是小鹏'三条曲线'战略。在他眼中汽车只是'四个轮子的机器人'，是技术演进中的一个节点而非终点。长期目标是用10年将小鹏打造为全球物理AI公司，海外收入占比达一半[s10]。

**Q9 (EN): Why is He Xiaopeng building flying cars and robots?**

A: He Xiaopeng believes there will be two kinds of giant technology companies in the future: fully digital companies, and global companies that combine hardware with software. Cars, flying cars and robots share the same physical-AI technology base, forming XPeng's "three curves" strategy. In his view, the car is merely "a robot on four wheels" — a node in the evolution of technology rather than its end point. His long-term goal is to build XPeng into a global physical-AI company within ten years, with overseas revenue making up half of the total [s10].

**Q10: 小鹏智驾XNGP与特斯拉FSD有什么区别？**

A: 特斯拉坚持纯视觉方案（摄像头+神经网络），不依赖激光雷达和高精地图；小鹏采用'视觉+激光雷达'融合感知方案，但也在向无图方向演进。小鹏图灵AI智驾（第二代VLA）采用720亿参数大模型、近1亿视频训练片段、每5天迭代一次，车端推理效率提升12倍。媒体实测800公里接管次数：特斯拉15次、华为2次、小鹏6次[s11]。

**Q10 (EN): What is the difference between XPeng's XNGP and Tesla's FSD?**

A: Tesla holds to a pure-vision approach (cameras plus neural networks), relying on neither lidar nor HD maps; XPeng uses a fused-perception scheme of "vision plus lidar," though it too is evolving toward map-free driving. XPeng's Turing AI driving (second-generation VLA) employs a 72-billion-parameter large model, nearly 100 million video training clips and an iteration cycle of once every five days, lifting on-vehicle inference efficiency twelvefold. In media road tests over 800 kilometers, takeover counts were 15 for Tesla, 2 for Huawei and 6 for XPeng [s11].

## 来源清单 Sources

1. [s1] 小鹏汽车 · 小鹏汽车官网·公司介绍与大事记 — https://www.xiaopeng.com/about.html
2. [s2] 小鹏汽车投资者关系 · 小鹏汽车投资者关系·董事会介绍 — https://ir.xiaopeng.com/zh-hans/governance/board-of-directors
3. [s3] 36氪 · UC创业十年：从邮箱到浏览器的逆袭 — https://m.36kr.com/p/1376373024961670
4. [s4] 环球企业家/CSDN · 环球企业家·UC优视的创业故事 — https://blog.csdn.net/jobding/article/details/83947903
5. [s5] 时代周报 · 阿里巴巴43.5亿美元收购UC深度调查 — https://www.time-weekly.com/wap-article/25282
6. [s6] 中国新闻网 · UC并入阿里幕后：马云的人性操作 — http://www.chinanews.com.cn/cj/2014/06-12/627181.shtml
7. [s7] 澎湃新闻 · 何小鹏：中年危机与造车梦 — https://m.thepaper.cn/baijiahao_9400308
8. [s8] 36氪 · 小鹏汽车十年深度报道 — https://m.36kr.com/p/3176053447856772
9. [s9] 创业邦 · 小鹏汽车产品发布与融资历程 — https://m.cyzone.cn/article/599859.html
10. [s10] 小鹏汽车投资者关系 · 小鹏汽车2025年Q4及全年财报 — https://ir.xiaopeng.com/zh-hant/node/9926/pdf
11. [s11] 太平洋汽车 · 小鹏图灵AI智驾（第二代VLA）技术解析 — http://m.pcauto.com.cn/hj/article/3370844.html
12. [s12] 汽车商业评论 · 汽车商业评论·何小鹏尖峰对话 — https://m-new.inabr.com/news/22722
13. [s13] 汽车商业评论 · 何小鹏做客《罗永浩的十字路口》 — https://inabr.com/news/21785
14. [s14] 晚点LatePost · 晚点LatePost·何小鹏CEO专访 — https://view.inews.qq.com/a/20250207A0522500
15. [s15] 凤凰网财经 · 凤凰网财经《封面》·何小鹏深度对话 — https://finance.ifeng.com/c/8oFVG2jpA3V
16. [s16] 南方周末 · 南方周末·何小鹏谈飞行汽车 — http://m.toutiao.com/group/7615920316132491822/
17. [s17] 新华网 · 小鹏汽车与滴滴战略合作公告 — http://www.news.cn/tech/20230828/dcba5257277444f5be52b58b70f4188f/c.html
18. [s18] 第一电动网 · 大众汽车向小鹏汽车增资7亿美元 — https://www.d1ev.com/kol/218521
19. [s19] 虎嗅网 · 小鹏集团Q2财报及机器人融资报道 — https://m.huxiu.com/article/4886513.html
20. [s20] 小鹏汽车 · 小鹏汽车2025年全年交付公告 — https://www.xiaopeng.com/news/company_news/5546.html
21. [s21] 国信证券/新浪财经 · 国信证券·小鹏汽车年报点评 — https://finance.sina.com.cn/wm/2026-04-09/doc-inhtvrts1230488.shtml
22. [s22] 小鹏汽车 · 小鹏汽车2026年全球新品发布会 — https://www.xiaopeng.com/news/company_news/5529.html
23. [s23] 小鹏汽车 · MONA M03上市及交付数据 — https://www.xiaopeng.com/news/company_news/5512.html
24. [s24] 飞行汽车官网 · 小鹏汇天飞行汽车产品发布 — https://www.feiauto.com/tt/1656.html
25. [s25] 铂道商业评论 · 何小鹏持股与投票权分析 — http://m.toutiao.com/group/7642191613124493867/
26. [s26] 小鹏汽车 · 小鹏图灵AI智驾发布会AMA直播 — https://www.xiaopeng.com/news/company_news/5544.html
27. [s27] 增长黑客 · 增长黑客·何小鹏管理风格转变深度分析 — https://www.growthhk.cn/cgo/model/133811.html
28. [s28] 新浪财经 · 小鹏与大众合作深化报道 — https://finance.sina.com.cn/stock/relnews/hk/2026-04-27/doc-inhvxhzn7921505.shtml
29. [s29] 百度百科 · 百度百科·何小鹏词条 — https://m.baike.com/wiki/%E4%BD%95%E5%B0%8F%E9%B9%8F/19330224
30. [s30] 民生周刊/人民网 · 民生周刊·何小鹏专访 — http://paper.people.com.cn/mszk/pc/content/202504/14/content_30069814.html
31. [s31] 小鹏汽车 · 王凤英正式加入小鹏汽车公告 — https://www.xiaopeng.com/news/company_news/4753.html
32. [s32] 汽车之家 · MONA L03 SUV发布 — https://chejiahao.m.autohome.com.cn/info/25792161
33. [s33] 小鹏汽车投资者关系 · 小鹏Q2财报与上半年业绩分析 — https://ir.xiaopeng.com/zh-hans/node/10176/pdf
34. [s34] 新浪财经 · 何小鹏2025胡润百富榜排名 — https://finance.sina.cn/stock/jdts/2025-11-20/detail-infxzray3784099.d.html
35. [s35] 上游财经 · 三个苦逼的忆苦思变——蔚小理三杰 — http://m.toutiao.com/group/7635670503642137130/
36. [s36] 南方人物周刊 · 南方人物周刊·何小鹏深度特写 — https://static.nfapp.southcn.com/content/202105/30/c5334666.html
37. [s37] 新浪财经 · 何小鹏接受央视《对话》栏目访谈 — https://finance.sina.cn/chanjing/gsxw/2025-02-07/detail-ineirwmy0584638.d.html
38. [s38] 汽车商业评论 · 汽车商业评论微博·何小鹏谈机器人 — https://m.weibo.cn/detail/5335606800679902
39. [s39] DoNews · DoNews·何小鹏双层玻璃办公室细节 — https://www.donews.com/news/detail/4/6021959.html
40. [s40] 证券之星 · 何小鹏2026胡润全球富豪榜 — https://cj.sina.com.cn/articles/view/7857141524/1d452771401902an1q

---
本档案机器版由亚洲封面人物官网自动生成：https://coverfigure.com ｜ 永久档案：https://coverfigure.com/acf/ACF-00-00108/geju ｜ 生成时间 Generated：2026-09-11 00:42