Peng Zhihui: The Huawei Engineer Who Became China's Robotics Folk Hero

In the West, the story of a robotics company usually starts with a founder who raises venture capital, hires researchers, and spends years in stealth mode. In China, the story of the world’s most prolific humanoid robot company starts with a man filming himself building machines in his apartment.

Peng Zhihui did not pitch investors before he had a product. He did not cultivate relationships with Sand Hill Road. He built a self-balancing bicycle in his living room, posted the video on Bilibili, and woke up to find that millions of people had watched it. He did this again. And again. And again. Over three years of building increasingly ambitious robotic systems on camera, Peng accumulated more than 10 million followers and became one of the most recognized engineers in China.

Then he left his job at Huawei, founded AgiBot, raised $140 million in his first year, and shipped 5,200 humanoid robots by early 2026, more than Tesla, Figure AI, and Apptronik combined.

This is not a startup origin story. This is a case study in how China’s maker culture, internet celebrity ecosystem, industrial supply chains, and government policy converge to produce outcomes that the West has no mechanism to replicate. And it starts with a camera, a soldering iron, and a young engineer who understood that showing your work is the most powerful fundraising tool ever invented.

The education of an engineer

Peng Zhihui was born in 1993 in Anhui province, a largely agricultural region in eastern China that is not typically associated with technology innovation. He studied automation engineering at the University of Electronic Science and Technology of China (UESTC) in Chengdu, one of China’s top engineering universities, where he completed both his undergraduate and graduate degrees.

What distinguished Peng from thousands of other capable Chinese engineering graduates was not his academic record, though it was strong. It was his compulsive need to build things and his equally compulsive need to share what he built. While still a graduate student, Peng began posting videos of his personal projects on Bilibili, the Chinese video platform that functions as a combination of YouTube and Reddit, with a user base that skews young, technically literate, and deeply engaged with long-form content.

His early videos were modest. A custom LED controller. A PCB design walkthrough. The kind of content that engineering students produce all over the world. But Peng had two qualities that transformed ordinary maker content into something much larger. First, he was an exceptionally clear communicator who could explain complex engineering decisions in language that non-specialists could follow. Second, and more importantly, his projects escalated in ambition at a rate that kept his growing audience in a state of constant anticipation.

A lighting controller became a custom flight controller for a drone. The drone project led to a self-balancing robot. The self-balancing robot evolved into a full robotic arm capable of precise manipulation. Each project built on the last, and each video documented not just the finished product but the entire process of design, fabrication, debugging, and iteration.

By the time Peng graduated, his Bilibili channel had already crossed one million followers. He was not a social media personality who happened to know engineering. He was an engineer whose work was so compelling that social media made him famous.

Huawei and the hardware discipline

In 2020, Peng joined Huawei as an engineer. The specific details of his role have never been publicly disclosed, but it is known that he worked within Huawei’s consumer electronics division, likely on hardware development for smart devices and related embedded systems.

Joining Huawei was significant for reasons that go beyond a line on a resume. Huawei is not just a large technology company. It is arguably the most disciplined hardware engineering organization in the world. The company employs over 200,000 people, invests more than $20 billion per year in research and development, and operates a corporate culture that is famously demanding, structured, and production-oriented.

For Peng, who had already demonstrated extraordinary individual talent, Huawei provided something he could not get from his apartment workshop: exposure to how hardware is built at scale. Not prototype hardware. Not demonstration hardware. Production hardware that ships in volumes of tens of millions of units per year, with reliability requirements measured in years of continuous operation and cost targets measured in fractions of a cent per component.

This experience would prove decisive. When Peng eventually founded AgiBot, he did not approach manufacturing the way most robotics startup founders do, as a problem to solve after the technology works. He approached it the way Huawei approaches everything, as the primary constraint around which every other decision must be organized.

During his time at Huawei, Peng continued posting to Bilibili. This was unusual. Huawei is a notoriously secretive company, and most employees keep low public profiles. But Peng’s content was personal, not corporate. He continued building and documenting his own projects, and the contrast between his day job at one of the world’s largest tech companies and his night-and-weekend life as a DIY robotics creator only made his channel more compelling.

His follower count continued to climb. By the time he left Huawei, Peng had more than 10 million Bilibili followers and was regularly cited in Chinese media as one of the country’s most influential technology figures. Not founders, not executives, not investors. An engineer.

The Bilibili videos that changed everything

To understand what made Peng’s content so powerful, it helps to examine the specific projects that went viral and transformed him from a popular creator into a cultural figure.

The self-balancing bicycle (2021). Peng built a bicycle that could balance and ride itself with no rider. The project required custom-designed control boards, inertial measurement units, motor controllers, and a sophisticated PID control algorithm. The video documenting the build accumulated over 20 million views on Bilibili and was widely shared on Weibo and WeChat. It was not just the technical achievement that captivated audiences. It was the sheer audacity of doing it as a side project.

The Iron Man helmet (2021). Peng designed and built a functioning Iron Man-style helmet with a motorized faceplate, heads-up display, voice recognition, and ambient lighting, all built from scratch. The project combined mechanical engineering, embedded systems, optics, and software development. The video crossed 30 million views and became one of the most-watched technology videos in Bilibili’s history. It demonstrated something important about Peng’s appeal: he could make advanced engineering feel like play.

The robotic arm with AI vision (2022). Peng built a 6-degree-of-freedom robotic arm with a camera-based vision system that could identify objects and execute grasping tasks autonomously. Unlike commercially available robot arms that cost tens of thousands of dollars, Peng’s version was built from commodity components for a fraction of the price. This project was the clearest foreshadowing of AgiBot. It showed that Peng was not just interested in making cool hardware. He was thinking about how to make robotic manipulation affordable and accessible.

The desktop humanoid (2023). In what many observers later recognized as a proof of concept for AgiBot, Peng built a small desktop humanoid robot with articulated limbs, camera-based perception, and basic manipulation capabilities. The video was accompanied by commentary about the future of humanoid robotics in industrial settings. Within weeks of posting it, Peng announced he was leaving Huawei.

The progression from hobby projects to a humanoid robot prototype was not accidental. Peng was building in public, and each project served a dual purpose. It entertained and educated his audience. It also proved to potential investors, partners, and future employees that he could execute, that he could ship hardware, and that he could attract attention for doing so.

The fame-to-factory pipeline

What Peng Zhihui built on Bilibili is something that has no direct equivalent in the Western technology ecosystem. In Silicon Valley, there is a well-understood path from Stanford or MIT to a startup, through accelerators, pitch competitions, and venture capital networks. In China, Peng demonstrated a different path: from engineering talent, through public demonstration of that talent on social media, to industrial company.

This pipeline is specific to China for several reasons.

Bilibili’s audience is the talent pool. Peng’s 10 million followers were not passive consumers. They were engineering students, working engineers, hardware designers, and technology enthusiasts. When Peng founded AgiBot, he did not need to post job listings on LinkedIn. His Bilibili channel was the most effective recruiting tool in Chinese robotics. AgiBot’s early engineering team was disproportionately composed of people who had followed Peng’s work for years and wanted to build with him.

Visibility accelerates fundraising. When Peng approached investors, he did not need to explain who he was or demonstrate that he could attract attention to a product. He had 10 million followers who had watched him build increasingly sophisticated robots. His track record was not a slide deck. It was a publicly viewable library of engineering accomplishments. Sequoia China, BYD, and SAIC Motor did not invest in a pitch. They invested in three years of documented capability.

Public building creates accountability. Every project Peng posted was a public commitment. His audience expected each project to be more ambitious than the last. This created a feedback loop that pushed Peng to tackle harder and harder problems, which in turn built the skills and the public record of achievement that made AgiBot possible.

Leaving Huawei, founding AgiBot

In early 2023, Peng Zhihui left Huawei and founded AgiBot, officially registered as Shanghai Zhiyuan Juren Technology Co., Ltd. The Chinese name, which translates roughly to “Intelligent Origin Giant,” reflects the company’s ambition. But the founding story reflects something more practical.

Peng has said in interviews with Chinese media that his decision to leave Huawei was driven by a specific observation: the gap between what humanoid robots could do in the lab and what they could do on a factory floor was not primarily a technology problem. It was a manufacturing problem. The components existed. The AI capabilities were advancing rapidly. What was missing was a company that could assemble those components into a reliable, affordable, mass-producible humanoid robot and get it into the hands of industrial customers fast enough to generate the real-world data needed to improve the next generation.

This was not an insight that would emerge from a research lab. It was an insight that came from building hardware on camera for three years and working at a company that ships hundreds of millions of devices per year. Peng understood tolerances, supply chains, component sourcing, and production line optimization. He understood that the robotics industry’s obsession with making the perfect robot was preventing anyone from making enough robots.

The founding team was small but deliberately chosen. Peng recruited engineers he knew from Huawei, from his university network, and from his Bilibili audience. The earliest hires were not AI researchers or computer scientists. They were mechanical engineers, manufacturing specialists, and supply chain managers. This hiring pattern was itself a statement of strategy. AgiBot was going to be a manufacturing company that used AI, not an AI company that happened to make robots.

The $140 million first year

AgiBot’s fundraising trajectory was extraordinary by any standard, but it was especially remarkable in the context of China’s venture capital market in 2023, which was experiencing a sharp contraction following years of regulatory crackdowns on technology companies and a broader economic slowdown.

The initial seed round came from Sequoia China (now HongShan) and the Shanghai AI Industry Investment Fund. These were not cold approaches. Sequoia China’s partners had been watching Peng’s Bilibili channel. The Shanghai AI fund was part of a broader municipal government effort to attract robotics companies to the city’s Lingang Special Area, the same industrial zone where Tesla’s Gigafactory Shanghai operates. The timing aligned Peng’s ambitions with Shanghai’s economic development priorities.

But the funding that defined AgiBot’s trajectory came from strategic investors, not financial ones. BYD invested in AgiBot not to generate venture returns but to secure early access to humanoid robots for its manufacturing operations. SAIC Motor, China’s largest state-owned automaker and a partner of both Volkswagen and General Motors in China, invested for the same reason. These companies needed humanoid robots on their factory floors, and they were willing to fund the company that could deliver them fastest.

The strategic investor model had a cascading effect. BYD’s involvement brought credibility that attracted other manufacturing companies. SAIC’s participation signaled government alignment. Sequoia China’s name brought international visibility. Each investor reduced the perceived risk for the next one.

More importantly, BYD and SAIC provided something no amount of money can buy: guaranteed deployment sites. AgiBot did not need to spend months or years convincing potential customers to try its robots. Its investors were its first customers, and they had factory floors ready to receive units as soon as production began.

This is the structural advantage that made everything else possible. American robotics startups raise money from venture capital firms and then spend years trying to find customers willing to take a risk on unproven technology. Peng raised money from the customers themselves. The difference in speed is not marginal. It is the difference between shipping 5,200 units and shipping 200.

Factory before perfection

The most consequential decision Peng made at AgiBot was also the most counterintuitive to anyone trained in the Silicon Valley model of startup development. He built the factory before the product was finished.

AgiBot established its manufacturing facility in Shanghai’s Lingang Special Area in 2024, less than a year after the company’s founding. The factory was not a prototype workshop or a small-batch assembly operation. It was designed from the outset for serial production of humanoid robots at volumes of thousands of units per year.

This decision reflected a philosophy that Peng articulated in a 2024 interview with 36Kr, one of China’s leading technology publications. He argued that in hardware, the factory is not the last step in the development process. It is the first step in the learning process. Every unit that comes off the production line teaches you something that no amount of simulation or prototyping can replicate. Tolerance issues that only appear at volume. Component failures that only manifest after hundreds of hours of operation. Assembly sequences that seem efficient in theory but create bottlenecks in practice.

By getting to production early, AgiBot compressed its learning cycle. The company’s first units were not perfect. They had higher failure rates, lower battery life, and less refined locomotion than what competitors were demonstrating in controlled environments. But they were real products deployed in real factories with real operational demands. Every problem was a data point, and every data point fed back into the next production run.

The approach has a direct parallel in how BYD built its electric vehicle business. BYD’s early EVs were widely criticized for inferior build quality compared to Tesla. But BYD shipped them anyway, learned from every unit, and used manufacturing volume to drive costs down to levels Tesla could not match. By 2024, BYD was selling more EVs than Tesla globally. Peng has explicitly cited BYD founder Wang Chuanfu as an influence, and the strategic resemblance is unmistakable.

What Peng’s company actually builds

AgiBot’s product line has evolved rapidly since the company’s founding, but the core philosophy has remained consistent: build robots that can do useful work today, not robots that demonstrate what might be possible tomorrow.

The initial products, the G1 and X2, were designed for specific industrial applications. The G1 is a compact humanoid platform optimized for environments where full-scale humanoid form is not necessary, while the X2 is a more capable platform for tasks requiring greater reach and payload capacity.

In early 2025, AgiBot launched the A2 series, which represents the company’s bid for the mainstream industrial humanoid market. The A2 comes in three variants: Standard, Max, and Ultra. All three run on WorkGPT, AgiBot’s proprietary multimodal AI system, deployed on NVIDIA Jetson Orin hardware.

The A2 Ultra, the top-tier model, stands 170 cm tall, weighs 75 kg, and can carry up to 20 kg. Its 42 degrees of freedom allow it to perform complex manipulation tasks in manufacturing and logistics environments. The unit price is approximately $100,000 for the Ultra, with the Standard variant priced significantly lower for simpler, more repetitive applications.

These are not the most technically impressive humanoid robots in the world. Figure AI’s 02 likely demonstrates superior AI-driven task generalization. Tesla’s Optimus Gen 3, with its neural architecture derived from Full Self-Driving, probably handles novel situations with more flexibility. But Peng was not trying to build the most impressive robot. He was trying to build the most deployable one. There is a difference, and it shows up in the shipment numbers.

The maker becomes the manufacturer

The most remarkable aspect of Peng Zhihui’s trajectory is not any single achievement. It is the speed and coherence of the transition from individual maker to industrial manufacturer. Each stage of his career fed directly into the next.

The Bilibili years gave him an audience that became his talent pipeline and his proof of capability. Huawei gave him the manufacturing discipline and supply chain knowledge that most startup founders spend years acquiring. The combination of public credibility and private expertise made fundraising fast. Strategic investors doubled as customers, collapsing the gap between funding and revenue. And the manufacturing-first approach turned early production into a learning engine that accelerated every subsequent improvement.

Each piece depends on the others. Remove the Bilibili fame and Peng is just another engineer with a business plan. Remove the Huawei experience and AgiBot probably follows the Western pattern of spending years on R&D before attempting manufacturing. Remove the strategic investors and the company has money but no customers. The full trajectory is what made AgiBot possible, and that trajectory is deeply specific to the Chinese technology ecosystem.

China’s celebrity engineer phenomenon

Peng Zhihui is the most prominent example of a broader phenomenon in Chinese technology culture. Unlike in the United States, where engineering talent is largely invisible to the general public and tech celebrity status is reserved for founders and executives, China has developed a culture that celebrates engineers as public figures.

This is partly a function of platform dynamics. Bilibili rewards detailed, long-form technical content in ways that YouTube and TikTok do not. A 45-minute video of someone designing a PCB from scratch can accumulate millions of views on Bilibili because the platform’s recommendation algorithm and user expectations support that kind of content. On YouTube, the same video would struggle to compete with shorter, more sensationalized formats.

But the phenomenon is also cultural. In Chinese public discourse, engineering ability carries a prestige that it does not carry in the United States. The most admired technology figures in China are often those who can demonstrably build things. Wang Xingxing of Unitree, who went from building robot dogs as a university project to running a publicly listed company, follows a similar pattern. Ren Zhengfei, the founder of Huawei, is revered in China not as a charismatic visionary but as an engineer who built the world’s largest telecommunications equipment company through relentless execution.

This cultural context matters because it creates a pipeline that the West does not have. When engineering talent is visible and celebrated, it attracts more people to engineering. When those engineers can build public profiles by demonstrating their skills, it creates a mechanism for identifying and funding the most capable among them. When the most capable engineers can raise capital from industrial partners who have seen their work, it creates a faster path from talent to company to production.

The result is companies like AgiBot, Unitree, and dozens of smaller Chinese robotics firms that were founded by engineers who built their reputations by building things in public. The pipeline is fast, efficient, and produces companies that are manufacturing-oriented from day one because their founders understand manufacturing from direct experience.

The open-source play

In 2025, AgiBot released the AgiBot World open dataset for embodied AI research, containing real-world robotic manipulation data collected from the company’s deployed fleet. The dataset covers grasping, navigation, object recognition, and task planning scenarios in actual industrial environments.

This was a Peng Zhihui move, and understanding why requires understanding his background as a maker and content creator.

Peng built his career on openness. His Bilibili videos did not just show finished products. They showed the entire process, including the failures, the debugging, and the wrong turns. His audience rewarded transparency with loyalty and attention. When he applied the same philosophy to AgiBot’s data, the results followed the same pattern.

The dataset also reflects a structural advantage that no amount of money can replicate. Open-source data from deployed robots operating in real factories, handling real materials, in real lighting conditions, is fundamentally different from simulated data. Companies without large deployed fleets cannot generate equivalent datasets regardless of their R&D budgets. AgiBot’s 5,200 deployed units are not just revenue sources. They are data collection platforms whose output improves every subsequent unit.

This is the flywheel that Peng designed. Ship units. Collect data. Release some data publicly to attract talent and set standards. Use all of the data, public and proprietary, to improve the next generation. Ship more units. Collect more data. The flywheel accelerates with every rotation, and each rotation widens the gap between AgiBot and companies that are still in the laboratory stage.

What Peng got right that Western founders got wrong

It would be unfair to characterize all Western robotics founders as having made the wrong decisions. They operate in a different ecosystem with different constraints, different capital structures, and different cultural expectations. But the comparison is instructive.

Peng prioritized manufacturing. Western founders prioritized technology. Brett Adcock at Figure AI, Elon Musk at Tesla, and Jeff Cardenas at Apptronik all spent years perfecting their robots before seriously addressing manufacturing. This approach makes sense when your investors are venture capitalists who reward technological breakthroughs and dramatic demonstrations. It does not make sense when the bottleneck to industry leadership is production volume.

Peng raised from customers. Western founders raised from financiers. AgiBot’s strategic investors were also its first customers. This collapsed the gap between fundraising and revenue in a way that venture-backed startups cannot replicate. Figure AI raised $1.85 billion in total funding, more than ten times what AgiBot raised, but had to independently find and convince every customer to deploy its robots.

Peng built his reputation before building his company. Western founders built their companies before building their reputations. This is perhaps the most underappreciated difference. When Peng founded AgiBot, 10 million people already trusted his engineering judgment. He did not need to prove himself. He needed to point that existing trust toward a specific product.

Peng embraced “good enough.” Western founders chased “best in class.” AgiBot’s robots are not the most technologically sophisticated humanoid robots in the world. Peng knows this. He does not care. His robots are good enough to do the jobs his customers need done today, and the revenue and data from those deployments fund the improvements that will make tomorrow’s robots better. The Western approach of waiting until the technology is exceptional before scaling production means that the data feedback loop never starts, or starts years later.

These are not moral judgments. They are strategic differences rooted in different ecosystems. But the outcomes speak clearly. In the time it took Figure AI to ship 200 robots, Peng’s company shipped 5,200. The question is not which approach is “better” in the abstract. It is which approach wins the market.

The government tailwind

Any honest account of Peng Zhihui’s success must include the role of Chinese government policy. Peng’s individual talent and strategic choices explain much of AgiBot’s trajectory, but they do not explain all of it.

In 2023, China’s Ministry of Industry and Information Technology (MIIT) released the Humanoid Robot Innovation and Development Guidelines, a policy document that established humanoid robotics as a national strategic priority. The guidelines called for China to achieve mass production of humanoid robots by 2025 and to establish global technological leadership in the field by 2027.

This was not a vague aspiration. It came with concrete support mechanisms. Municipal governments, including Shanghai’s, offered subsidies, tax incentives, and preferential access to industrial zones for robotics companies. The Lingang Special Area, where AgiBot built its factory, is a government-designated zone with reduced corporate tax rates, streamlined permitting, and access to shared infrastructure.

The government’s role was more structural than financial. By declaring humanoid robotics a national priority, MIIT created a permission structure that encouraged state-owned enterprises and large private companies to invest in and deploy humanoid robots. When BYD and SAIC invested in AgiBot, they were making a business decision. But it was a business decision that aligned with government priorities, which reduced the perceived risk and increased the speed of deployment.

This alignment between individual ambition, corporate strategy, and government policy is the Chinese technology ecosystem’s distinctive advantage. It is not central planning in the Soviet sense. It is coordinated capitalism, where government policy shapes market incentives in ways that accelerate specific outcomes. Peng did not need the government to tell him what to build. He needed the government to create an environment where building it quickly was possible and rewarded.

What Peng’s story reveals

Peng Zhihui’s trajectory from Bilibili creator to Huawei engineer to humanoid robot manufacturer reveals something important about how technological leadership is built in the 2020s.

It is not built by having the best technology. Tesla and Figure AI probably have more sophisticated AI systems than AgiBot. It is not built by raising the most money. Figure AI has raised more than ten times what AgiBot has. It is not built by hiring the most researchers or publishing the most papers or giving the most impressive demos.

Technological leadership is built by shipping products. By getting hardware into the hands of customers. By collecting real-world data from real-world deployments. By using that data to improve the next generation of products. By manufacturing at a cost and speed that competitors cannot match. And by doing all of this before anyone else does.

Peng understood this because he spent three years building things on camera and learning what it takes to make something that works. He understood it because he spent two years at Huawei learning how hardware is built at scale. And he understood it because he grew up in a country where the path from technical talent to industrial production has been optimized by decades of experience in solar panels, batteries, drones, electric vehicles, and telecommunications equipment.

The West has no equivalent pipeline. There is no American Bilibili where engineers build massive audiences by showing their work. There is no American equivalent of the BYD-SAIC model where industrial companies fund robotics startups and immediately deploy their products. There is no American equivalent of the MIIT guidelines that coordinate government policy, corporate strategy, and startup ambition toward a unified industrial goal.

This does not mean the West cannot compete. Tesla’s resources, Figure AI’s AI capabilities, and the deep reservoir of American robotics talent represent genuine strengths. But competing requires first understanding what you are competing against. And what you are competing against is not a company. It is a system.

Peng Zhihui is the human face of that system. A maker who became a celebrity who became an engineer who became a manufacturer. His story is singular, but the infrastructure that made it possible is not. And that infrastructure is producing more humanoid robots, faster and cheaper, than anything the Western world has built.

The cameras are still rolling on Bilibili. Somewhere in Shanghai, the factory lines are running. Both of those things are true at the same time. That is the point.