In 2014, He Xiaopeng (何小鹏) became financially free overnight when Alibaba bought his mobile-browser company for USD 4.35 billion. Three years later he left Alibaba, took over a struggling electric-car startup, and began telling anyone who would listen that the real prize was not the car at all. Today he runs XPeng (小鹏汽车) as a Group building three physical machines at once: AI cars, flying cars, and humanoid robots.
Few Chinese founders have staked so much of one balance sheet on embodied intelligence (具身智能) across three very different product classes.
The serial builder
He was born in 1977 in Huangshi, Hubei, and graduated from South China University of Technology (华南理工大学) in 1999 with a computer-science degree. In 2004 he co-founded UCWeb (UC优视) with Liang Jie and Yu Yongfu; its UC Browser became a dominant mobile gateway in China's early smartphone era. In June 2014 Alibaba acquired UCWeb for about RMB 43.5 billion (USD 4.35 billion), then the largest internet acquisition in China, and He joined Alibaba as president of its mobile business group.
He had already backed XPeng, founded in 2014 by Xia Heng and He Tao, and in August 2017 he left Alibaba to become chairman and CEO. Under him XPeng listed on the New York Stock Exchange in August 2020 and added a Hong Kong dual listing in 2021, riding China's premium-EV wave.
Why "AI-defined"
He's central thesis is that a car is increasingly a robot on wheels, and the same AI stack — perception, planning, end-to-end models — transfers to other physical forms. XPeng now describes itself as building "physical AI" (物理AI) across three platforms: AI vehicles, flying cars, and AI robots.
The anchor product remains the car, but the differentiator he pushes is autonomous driving. XPeng's smart-driving system, built on a large end-to-end model and trained in a dedicated cloud compute cluster, is the capability he believes will spread to the other two platforms.
IRON, the humanoid
At XPeng's AI Tech Day on 6 November 2024, He unveiled IRON, a humanoid (人形机器人) built in-house. Its stated specifications:
- Height 1.78 meters, weight 70 kg, with 62 body degrees of freedom and 15 degrees in each hand.
- A self-developed Turing AI chip quoted at 3,000 TOPS of compute, running an end-to-end model with reinforcement learning.
- A 720-degree "Eagle Eye" vision system borrowed from XPeng's driving stack.
- Hands with tactile feedback for object manipulation.
Crucially, IRON was not shown only on a stage. He said it was already working inside XPeng's Guangzhou factory, performing tasks such as assisting assembly of the P7+ sedan. In 2025 XPeng showed an updated IRON with three Turing chips, a solid-state battery, and a physical-world AI model, and targeted mass production in 2026. He has said the goal is an early L3-class humanoid capable of real factory and eventually service work.
The flying car
The second non-car bet runs through XPeng AeroHT (小鹏汇天), its aviation arm. Its "Land Aircraft Carrier" (陆地航母) is a split-body flying car: a six-wheeled ground vehicle that carries a foldable two-seat eVTOL aircraft in its rear bay, deployable for short low-altitude hops. Reporting from IFA 2026 put pre-orders near 7,000, mostly from China and the Middle East, with mass production targeted for the second half of 2026, an initial capacity around 5,000 units a year, and a price near USD 280,000. China's civil aviation authority (CAAC) has granted a foundational manufacturer-suitability certification, with vehicle approval expected to follow.
Taking the robot business personally
In 2026 He tightened his grip on the robotics push. In June he announced he would directly lead XPeng's robotics unit and became its business CEO; in August the robotics arm closed a first funding round reported above USD 900 million. At China's 2026 "Two Sessions" he proposed building an intelligent-standard system for humanoid robots, signaling he wants policymakers alongside him.
Honest limitations
- IRON is pre-mass-production. Despite factory trials, a humanoid doing reliable, safe, all-day physical work at scale is not yet a shipped product; 2026 targets are intentions, not deliveries.
- Flying cars face airspace, not just engineering. Certification, pilot licensing, and urban integration remain unresolved; the USD 280,000 price puts it far from mainstream use.
- Three fronts, one P&L. Pouring capital into EVs, humanoids, and eVTOLs simultaneously tests even a well-funded group, and humanoid and flying-car lines will lose money for years.
- Self-reported specs. Chip TOPS figures and deployment counts come from XPeng's own stage presentations and should be treated as claims until independently verified.
What readers can do now
- Watch factory deployment, not launch videos. IRON's real test is whether XPeng keeps it working inside its own plants through 2026 — that is harder evidence than a keynote.
- Track the robotics funding round. A reported USD 900M raise shows outside belief; follow whether the unit hits mass-production milestones on schedule.
- Follow the standards push. He's lobbying for humanoid-robot standards; that regulatory work may shape China's entire embodied-AI (具身智能) sector, not just XPeng.