Hi all,

We have extensively covered how China dominates the hardware “body” of embodied AI through cheap, vertically integrated robot manufacturing; however, we haven’t compared it to how it stacks up against the U.S., which leads the “brain” side with cutting-edge model architectures and disciplined data practices.

Embodied AI is the fusion of metal and mathematics, as Jefferies’ Shenzhen expert reminds us in a recently published report, JEF Industrial AI+ Value Chain Expert Day: Embodied AI Model. Based on the interviewed research scientist from the Chinese University of Hong Kong (Shenzhen campus), whoever cracks this integration first will set the cost curve for $20K “robot co-workers” capable of reshaping global labor economics. I continue not to like the framework of China vs. the U.S., but there is some reality in which the market is fundamentally more capable in certain areas. Thus, we can see that this race is no longer theoretical—purchase orders are already crossing the Pacific.

Brains vs. Brawn

We’ve covered extensively where China is doing well in embodied AI, but there is a caveat, which is in the manufacturing process. Beijing’s advantage is tangible. It is a comprehensive humanoid-robot supply chain that lets startups source everything from harmonic reducers to vision sensors without leaving the Pearl River Delta. Factory pilots scale fast because unit economics resemble appliances more than moonshots. Domestic capital, hungry for quick returns, pushes teams to ship hardware first and software later, and hence the parade of slick chassis at Shenzhen roadshows.

But this hardware obsession leaves the “mind” or “brain,” whatever you’d like to call it, still vastly underfunded. Data creation, collection, and management remain patchy, costly, and complex, as confirmed by the UBtech executive. Tesla benchmarks “over $100 per data point,” according to industry insiders, and inconsistent labeling standards keep China behind the U.S. in both the quality and quantity of real-world data.

Across the ocean, Silicon Valley dominates the invisible stack. The U.S. excels in scientific research, technological innovation, and capital-driven growth but struggles with owning the humanoid supply chain and real-world application scenarios, according to think tank RAND. Expertise in CUDA programming, simulation farms, and multi-sensor data pipelines gives Google’s RT family and Tesla’s Optimus an algorithmic lead. However, every lightweight gripper they bolt on still ships in crates stamped “Made in Suzhou” (also known as the Venice of China).

The Catch-up Calculus

Now, back to the Jefferies report, it estimates China lags the U.S. by roughly ONLY a month in pure large language model (LLM) capability now. The interviewed expert said that if the U.S. opensources a full embodied AI framework, Chinese labs could match it within three months, assuming open weights, affordable compute, and unhindered GPU imports, all variables Washington can still influence.

And there is a reason why China seems to have gone all open source; in many ways, it’s been driven by the idea that it’s better for China to catch up first, no matter who it is, as long as it can propel China’s AI sector forward.

On the hardware front, in the near term, let’s say the 6-12 month horizon, it seems like there will continue to be a rising demand for China's actuators, reducers, and lithium-packs. However, rising export inquiries are already lengthening lead times, and we’re beginning to see an early indication that demand is outstripping capacity and pricing power is shifting in their favor. That means, without a resilient, China-independent supply chain, U.S. pure-play robot-hardware companies may face some challenges. How the tariffs and export controls continue to play out is a question no one can answer at this point.

Deep dive on South Korea’s AI ecosystem with an exclusive interview with “Korea’s Sequoia” coming up.

Today, I want to introduce a dear friend, Nico, who will be contributing to AI Proem regularly. Nico Finn provides tech insights based on fieldwork and data. She holds an MPhil from Oxford and is currently a Yenching Scholar at Peking University.

She has written extensively on China’s robotics, AI, and tech sectors widely, and this insightful report elaborates further on the companies to watch and the practical use-cases where physical AI is being used already. See her recently published story on China’s robotics industry here.

Subscribe now

(And passing it to Nico, mostly…)

China, U.S., humanity vs. Robots

My recent trip to China showed me numerous robots deployed in real life. Seeing dish-delivering robots at every hotpot restaurant and USB cord charger delivery robots at hotels is not exactly jaw-dropping, but it certainly invites my fellow American travelers to rethink the scale of China’s market appetite for robotics and embodied AI at large.

Embodied intelligence technologies are claiming a growing share within this expanding market. Several structural advantages underlie China’s momentum in this field. According to the 2023 China Robotics Industry Development Report, China’s service robot market exceeded RMB 50 billion (~USD 7 billion) with annual growth surpassing 30%. As discussions about the AI competition, while much attention has been paid to foundation models, large language models (LLMs), and general-purpose AI capabilities, the arena of embodied intelligence presents a more nuanced and salient landscape. China has rapidly developed distinctive advantages in this domain that challenge prevailing assumptions of China’s cheap/ not as advanced technology, especially compared to the U.S.

On that note, Jia Yueting, the former CEO of LeEco, shocked everyone by appearing at BEYOND Expo at the opening ceremony. He seems to be another former executive who is unretiring due to AI (like Sergey Brin).

After many years of living abroad, he brings a more nuanced perspective on China vs. the U.S. Jia Yueting said, "Many people believe that AI in the US and China is in opposition and mutually exclusive. But I have always believed that AI is this era's most powerful transformative medium—it inherently possesses the attributes of integration, reconstruction, and coexistence. AI in the U.S. and China is not an either-or situation; it is entirely possible to achieve complementary win-win outcomes, promoting a broader and more orderly release of AI's value on a global scale."

"Looking at the world from the perspective of the U.S., the four new routes for Chinese enterprises to go global in the future are the bridge model, the overseas expansion of industries and supply chains, the overseas expansion of AI ecosystems, and the overseas expansion of Web3 ecosystems. The boundary-breaking transformations of these four elements will also become a significant trend in reshaping the global business landscape," said Jia Yueting.

This primarily aligns with what many startup founders are saying in China and the businesses and services we are seeing being exported. This generation of entrepreneurs truly believes that going global cannot and will not be stopped by geopolitics, especially given the sophistication of founders nowadays.

What is embodied intelligence? It goes beyond humanoid

Embodied intelligence refers to the confluence of AI with robotic systems, where intelligence is instantiated not just in data processing but in physical form. This includes, of course, humanoid robots, but also service robots, drones, autonomous mobile platforms, and intelligent manufacturing systems. On the flip side, these robots must be “smart” and have self-upgrading capacity to be considered embodied intelligence. Many firms and labs now focus on combining the likes of LLMs and robots to make them learn intelligently. Unlike LLMs, which primarily inhabit digital spaces, embodied intelligence allows applications to extend into labor markets, eldercare, education, logistics, and military scenarios. This extension into the real world means the stakes are not just about data but about material interactions, infrastructure, and human well-being. [Grace has previously shared some insights that the UBtech COO shared recently on physical data and physical intelligence here]

China has many structural advantages. And China knows it

This is evident in this year’s Spring Festival Gala, where China usually broadcasts its most note-worthy technological achievement to the world. Previously, aerospace missions and high-speed railroads had all been featured, nearly like a national rite of passage. This year’s Unitree H1 humanoid robot was the protagonist. It is the strongest visual evidence echoing what many believed in China, that 2025 is ‘the year 0 of the epoch of Chinese robotics’.

With a population of 1.2 billion+ and the most complete manufacturing ecosystem globally, China offers a unique environment for rapid prototyping and scale. China has developed a highly synergistic ecosystem, or “integrated industry policy”, from core components such as Inovance’s servo motors to integrators like Siasun Robotics. For instance, DJI’s technologies have contributed to motion control and sensor fusion capabilities crucial to embodied intelligence, not just its final products in the form of drones. [Grace has written bout the rise of industrial robotics here]

Taking a step back from products, patent records also show that China has been poised to provide more breakthroughs for embodied AI. The World Intellectual Property Organization reports that China has led global generative AI patent applications for four consecutive years, and China now boasts more than 190,000 effective robot-related patents, or ⅔ of the global total.

So, what are some scenarios using robots?

This goes beyond the showy robots like the dancing humanoid and food-delivery drones. The following scenarios are a lot more low-key, but are exactly going to be the backbone of the industry’s success.

Advanced manufacturing

The Chinese manufacturing sector is embracing collaborative robots, or “cobots,” to upscale its economy. These robots work alongside human workers, handling tasks such as welding, assembly, and quality inspection. This was largely showcased in the UBtech x Zeekr March demo as well. Where robots worked together with each other and with humans on the production line.

Another company worth mentioning is one of the largest manufacturers, Siasun, which launched DUCO GCR cobots designed for welding and assembly. Estun, another company, specializes in spot welding and arc welding, targeting particularly the automotive industry.

Hikrobot is a subsidiary of Hikvision, known to have the world’s most advanced facial recognition products, and develops mobile robots and machine vision products for intelligent intralogistics solutions in industrial settings. The company will be known beyond its world-leading security cam, but also for advanced manufacturing and cobots, ranging from a handheld code reader, vision box, lens, light source, and more.

Agritech

Agritech actually has a significant market application potential in China and beyond. DJI, DJDynamics, XAG have been providing numerous leading models of drones and integrated smart systems for precision crop spraying, field monitoring, and fruit picking. The DJI Agriculture 2022 White Paper report highlighted that already 210,000 drones were in operation globally as of 2022, while DJI Agriculture has already expanded to over 100 countries to offer tailored drone solutions to local crop and terrain conditions.

In the next few years when you take a tequila shot, it’s likely the hard labor and fruition of one of these DJI drones. (Not to mention that the lychee for your cocktail could be well harvested by robots in southern China as early as 2022.)

Healthcare

Healthcare is another big category of AI with a readily available market demand. Robots are being deployed for surgical assistance, hospital logistics, and rehabilitation. MicroPort MedBot’s laparoscopic surgical robot has received market approval in China, facilitating minimally invasive surgeries across departments. Fourier offers solutions to aid physical therapy and recovery, which is an obvious demand for robotic assistance. TMiRob disinfection robots utilize UVC light and hydrogen peroxide mist to sterilize hospital environments, enhancing infection control.

Delivery and logistics

Like the hotpot ingredient-delivery robots that surprised our American travelers in China, many service and hospitality robots are excelling at logistics and customer interactions. Pudu is a smart delivery robot designed for restaurants, and Yunji offers “room service” delivery to hotel guests - also something very common in China nowadays, even at a random four-star hotel that I recently visited. Others now sell humanoid robots for reception and concierge roles, integrating AI to interact with customers in service settings.

Home appliances

Smart home appliances are not unfamiliar terrain for many of us. I remember getting a Roomba vacuum cleaner at Costco ten years ago. But a decent first-hand Roomba costs between USD 450 and USD 1000, while a reputable Chinese brand would cost only half of this. And this is what China will continue to thrive on. Roborock, another Chinese company that sources its manufacturing with Xiaomi, is now a top product on U.S. Amazon. Roborock boasts to have the best mapping features on the market. Its latest model even has an arm to pick up small obstacles like a sock…Currently, the sock-picking vacuum costs over a grand, but when the day comes when it is competitively priced, it could be the end of Roomba.

Parting words, for now

In terms of defining what constitutes ‘embodied AI’ tech, scholars and researchers are wary of conceptual stretching. We think that just like autonomous driving, the “real embodied intelligence” is not here yet. But here, like Grace Shao at AI Proem, I also echo Dr. Jeffrey Ding’s thesis that to win in the tech realm, sometimes it is not about having the most cutting-edge tech but the scale of adaptation and leveraging the opportunity to upscale the whole economy.

From what I gathered on the ground, this is what I see: Chinese companies and policymakers are extremely good at. Electric vehicles are one such example. Although Tesla was the flagship of EVs worldwide, Chinese EVS - both big and small brands- have come to dominate actual sales globally, especially for markets outside of the U.S. Chinese professionals working in the robotics field recognize that the U.S. will likely continue to have an edge on cutting-edge tech. Still, China is competitively positioned to become one of the largest producers of advanced robots.

We’ve covered embodied AI extensively recently, despite the confidence being garnered by the flashy robotic demos and talks by the likes of Unitree, Galbot, Tesla, and Nvidia. There are still real challenges for embodied AI to become a reality for any of us anytime soon.

The lack of physical data remains the biggest bottleneck for embodied AI models: Embodied AI models require diverse datasets to generalize to new, unseen tasks. With LLMs, text, image, and video data are mainly available on the internet, but for robotics models, companies and research labs are creating new 3D data. The issue is that the necessary sensor and actuation data do not exist on a scale that can train generalizable AI models yet. And this was echoed by UBtech’s CBO, Michael Tam, at the Beyond Expo.

As of where technical development is at now, LLMs can't go so far in the physical world yet. The robots' comprehension of language-directed instructions is quite limited. The nuanced movements required even to lift a pen, such as the angle, the force with which to hold it, and the temperature at which to touch the item, as well as the height at which to lift it, are too complex for robots to understand. There are over 50 precision actuators in a humanoid robot designed for full-body tasks, and that is the kind we see today, which still appears somewhat bulky. Humans have about 360 joints throughout our bodies. Now imagine how hard it is to require the robots to move “as swift as a coursing river, with all the force of a great Typhoon” (yes, I’m quoting Mulan).

Subscribe now

3 Primary Ways to Train a Robot

So, how are companies overcoming these challenges or at least trying to improve the training of these robots? We’re seeing that there are three primary approaches taken by companies in training physical AI models, according to Morgan Stanley’s research.

1. Teleoperation: This is the most traditional method of training robots, where a human operator uses a remote control to guide the robot through a series of movements to complete a specific task. And while the robot is being manipulated, its range of sensors can capture the vision and physical data of the series of movements, and then use that data set to train its neural networks underlying the model for the robot. This will be repeated thousands of times with slight tweaks and modifications based on the surrounding environment and specific task, so finally, in time, the robot will be able to take command and complete the action in similar scenarios. The fact is, this type of training can be highly time-consuming and labor-intensive, but it is the most effective way to train physical data.

2. Simulation: This option has infinite possibilities, as simulations can be diverse in setups. Humanoid robots have a higher number of degrees of freedom (DoF). And thus, they can be much more “free-spirited” in how they complete tasks in more unpredictable environments. But this could be much more computationally intensive.

3. Real-World Videos: Often, videos of humans or robots performing tasks and actions are used as training data, primarily for training robots to operate on assembly lines. The data can be vast, as we have a lot of available material online, but this again does not solve the issue with how accurately the movements can be translated to physical data, since it's only visual.

In Reality

What most embodied AI companies are trying to do now is use a hybrid training version of the three mentioned above.

Companies like Nvidia are already offering synthetic data training for robots, which enables them to create simulations, generate physical data, apply data augmentation, and store that data for companies to train their robots on.

But as we keep going back to the point that humanoid robots are not just a hype, but then for any model to really reach the average household or find a consumer use case at scale, may still take 5-10 years, or even more.

And even tech optimists are holding reservations on this front. Here is an excerpt I gathered from the Beyond Expo in Macau last week, specifically regarding robotics by Joe Tsai, Chairman of Alibaba.

Tsai was the keynote speaker for the closing ceremony, and he emphasized that when artificial intelligence is infused into robots, existing robots will become smarter and even capable of thinking. Robots can move in a way driven by reasoning and thinking, rather than just executing tasks according to pre-written programs.

This technology is not limited to factory automation but has a much broader range of applications. For example, robots can sweep the floor, make coffee, and play musical instruments, all of which are truly amazing.

On the other hand, when it comes to the utility of robots, most intelligent robots in the world do not need to look like humans - which I also agree with, even though UBtech has said humanoid robots will be the best to integrate into this physical world that is designed for us.

He even joked that he might feel scared if his robot cleaner looked like a human and would prefer it to look like a vacuum cleaner, but be able to move intelligently around the room.

Currently, most use cases for robots or intelligent robots are applicable. However, when it comes to humanoid robots, the real consideration is for those scenarios where people want something that looks and feels human, and there are not that many such scenarios. In these cases, I would probably prefer to interact with a real person rather than a robot.

Although artificial intelligence can now understand knowledge and reason, there is still a need for more progress in spatial intelligence, which is a challenge that the robotics industry needs to overcome in order to move forward.

Share

Some Comments on Talent and Globalization

A recent conversation with a Silicon Valley-based investor highlighted an interesting trend that is quietly unfolding. There are currently numerous restrictions on capital injection from the U.S. into the Chinese AI market. - that we know. HOWEVER, there are a handful of U.S. AI companies that are operating in stealth mode in China.

I was initially perplexed, as the AI market is quite saturated with domestic players that are often more cost-effective for users in China. Furthermore, considering the sensitivities, I didn't see an apparent business reason for U.S. AI companies to do so. But the investor pointed out a reason I overlooked, which is the talent pool.

For a U.S. company to set up operations in China means they can easily tap into the local talent pool without paying the crazy Silicon Valley compensations (supposedly USD 1.3 million+ TC for a junior AI developer, wowza) and MUCH more for more senior talent. Now, setting up shop in China means that these companies can access some of the top researchers, developers, and engineers in the world right now, and without disrupting the lifestyles of these talents, who likely don't want to move to the U.S.

What we can predict is that this trend may even be exacerbated by the new Trump push to rescind student visas for Chinese students in sensitive sectors or those deemed to have political ties. As more will be forced or choose to move their studies and research back to the Mainland, or likely, I think, Hong Kong and Singapore, where STEM institutions rank highly globally.

In reverse, Chinese AI and tech companies continue to seek opportunities to “go global,” with the U.S. market being the ultimate goal for many. And although SHEIN’s IPO plans shifted from New York to London and now to Hong Kong, there is still hope for many to access global investors.

For now, stay tuned. Expect a deep dive and a list of companies that are finding real-life use cases for robots in healthcare, agritech, and manufacturing next Tuesday at ~ 7:30 AM EST from our special contributor, .

Related analyses:

• Rise of China's Robotics Industry: from Manufacturing Arms to Embodied AI

• Electric Dreams, Robot Reality: China's EV-to-AI Evolution

• Physical AI’s ChatGPT “moment”: A Closer Look at Embodied AI Robots

• Robot Girlfriends, Robot Firefighters, Robot Dogs, Robot Maids and Beyond

• Rise of China's Robotics Industry: from Manufacturing Arms to Embodied AI

• Introducing Unitree, China’s leading AI-embodied Robotics Company

Now, a belated but long deep dive, with the support from my “best friends” Manus and ChatGPT o3.

China's robotics industry did not just take off overnight; it has been years, if not decades, in the making. As I’ve mentioned in the EV to Robotics piece, many of the tangentially related industries grew together, propelled by national planning and policies and technological and manufacturing synergy.

Subscribe now

In recent months, with the talk of reshoring manufacturing jobs to the U.S., attention has been directed to the possibility of using industrial robots to fill in the gaps, and a discussion of whether China has been utilizing robots to replace some of the aging labor force.

Since the rise of globalization, China has been known to be the manufacturer of all goods, essentially for the world, including everything from smartphones to car parts to home appliances.

Today, we explore how that has laid the foundation for some of its advanced physical AI. More so, has China been adopting industrial robots and transitioning from human labor to robotic labor, especially given the country’s declining workforce and aging population?

The robotics industry has undergone a remarkable transformation over the past five decades, evolving from basic research and prototype development to becoming a global leader in industrial and embodied AI robotics. This comprehensive report examines the origins, development trajectory, technological strengths, applications, government policies, and entrepreneurial ecosystem shaping China's robotics sector.

A. Long-term Strategic Planning

Notable Milestones in the Recent 10 Years

• 2015: Launch of "Made in China 2025" initiative, identifying robotics as a strategic industry

• 2016: China's first National Robotics Development Plan released

• 2021: Release of the 14th Five-Year Plan for the Robotics Industry

• 2023: Introduction of the "Robotics+" Application Action Plan

• 2023: Release of Guiding Opinions on the Innovative Development of Humanoid Robots

• 2025: First mention of embodied AI in the Government Work Report, signaling its strategic importance

China has ~1.75 million industrial robots on factory floors, more than any other country, and 51% of global annual installations as of 2023. According to a Siemens-sponsored robotics podcast, the Germans are also adopting more across industries, but are still nowhere close to the adoption rate. This is also mainly due to the cost efficiency advantages of the products in China. The cost of a six-axis arm has fallen more than 60% in the last decade, and that is due to the hardware manufacturing engineering and know-how that we’ll go into further below as well. (Below, we go into details of the most important policies that drove this development)

B. Early Development of China's Robotics Industry

Infancy Stage (1970s-2010s) China's robotics industry has evolved through several distinct phases

• Initial Research Phase (1970s-1980s): China's robotics journey began in the 1970s with fundamental research at institutions like the Chinese Academy of Sciences (CAS) and Harbin Institute of Technology (in the very far North East of China). Harbin has a comprehensive industrial system that manufactures power equipment, aircraft, electronics, chemicals, and more. Harbin Electric Company Limited is one of the three largest manufacturers of power plant equipment in China, producing steam turbines and generators. So it also makes sense that the city nurtured the first Chinese industrial robot prototype, the "JR-1," which was developed in 1979 at the Shenyang Institute of Automation under CAS.

  You may have heard of Harbin for tourism reasons, as it has a famous annual ice sculpture festival that attracts millions of visitors annually. But I’ve never been, as I’ve been “wintered out” after my four years in Edmonton. If interested, you can look into the HIT Robot Group, a JV between a private company, the local provincial government, and the Harbin Institute of Technology, which specializes in industrial robotics and intelligent equipment.

• Prototype Development Phase (1990-2000): During this period, China focused on developing indigenous robot prototypes while beginning to import industrial robots from international manufacturers like ABB, KUKA, and FANUC. Many of these mentioned were European-homegrown and are now bought out (or attempted to be bought out) by Chinese firms. Research institutes and universities led most development efforts, with limited commercial applications.

• Early Commercialization Phase (Late 1990s-2000): The late 1990s saw the emergence of China's first robotics companies, including Siasun Robot & Automation Co. (founded in 2000), which grew out of the Shenyang Institute of Automation. These early companies primarily focused on industrial automation applications.

• Mass Industrialization Phase (2001-2010): As China's manufacturing sector expanded rapidly, the demand for industrial robots grew substantially. This period saw increased adoption of robotics in automotive manufacturing, electronics assembly, and other industrial applications. However, most robots were still imported from foreign manufacturers.

• Ultra-Rapid Development Phase (2011-Present): Since 2011, China's robotics industry has experienced explosive growth. In 2013, China became the world's largest market for industrial robots, surpassing Japan and South Korea. The government's "Made in China 2025" initiative, launched in 2015, identified robotics as a key strategic industry, accelerating development through policy support, funding, and research initiatives.

Prior to the 2010s, China lagged significantly behind established robotics powers like Japan, Germany, and the United States in core technologies such as precision components, control systems, and software. In most industrial areas, it largely depended on imports, and imported machinery was seen as much more premium in quality and precision than the ones made domestically. That changed in the mid-2010s.

C. Manufacturing Ecosystem: Scale & Vertical Integration

However, China's manufacturing prowess comes from grand infrastructure, cheaper labor, and decades of practice. The nation’s manufacturing capabilities, whether in renewables, automobiles, textiles, or industrial hardware, give it an advantage in outpacing the U.S. in production if necessary.

China’s edge in robotics begins where most machines are born: the workshop itself. Decades of building everything from textiles to EV batteries have woven a dense industrial fabric—suppliers, component makers, tool-and-die shops—often within a short truck ride of one another. That proximity lets robot start-ups whip through prototyping cycles in days, not months, while established players tap the same network to ramp production fast. Many firms keep design, gearbox machining, servo assembly, and final testing under one roof, squeezing out middle-man mark-ups and tightening quality control. The result is that a six-axis arm or a bipedal prototype can roll off the line at 30-50 percent less than a comparable Western build—an advantage that helps explain why China installs more factory robots each year than the following four countries combined.

Batteries Powering the Future
A robot is only as helpful as its last charge, and China owns roughly three-quarters of the world’s battery-cell capacity. Know-how transferred from the electric-vehicle boom is now reshaping robotics. Higher-density chemistries stretch run-times; flexible, custom-shaped packs free up designers to slim torsos and limbs; and sheer manufacturing scale keeps costs on a downward glide path. Equally important are innovative battery-management systems—often written by the same teams that once tuned EV packs—that monitor temperature and load in real time, squeezing out extra cycles and boosting safety on the factory floor.

Sensors, Actuators and AI: Eyes, Muscles and Brains
Low-cost depth cameras, LiDAR units, and environmental sensors—many sourced domestically—feed robots a richer data stream without blowing the bill of materials. Paired with lighter, more precise servo motors and linear actuators, these “eyes” and “muscles” translate into smoother pick-and-place moves or steadier bipedal gaits. Layer on China’s fast-evolving AI stack—edge chips for perception, cloud back-ends for continuous learning—and robots improve in the field, not just on the drawing board. Supportive technologies such as 5G and industrial-grade IoT networks round out the package, letting fleets report status or receive new motion policies over the air.

Supply-Chain Depth: From Rare Earths to Final Freight
Behind the glossier tech lies a less visible but critical moat: materials and logistics, which are now also at the forefront of the trade discussions between China and the U.S.

China controls about 60 percent of global rare-earth output, giving domestic motor and sensor makers privileged access to neodymium and dysprosium. Specialty semiconductor fabs—while still catching up at the bleeding edge—have already pumped out the motor drivers and AI accelerators that robots require. With component vendors, contract manufacturers, and ports clustered along the same coastal corridors, lead times shrink and inventories stay lean. The concentration also cushions shocks: if one supplier stumbles, another is often a city, or sometimes a city block, away. For robot builders racing to hit price points and production quotas, that resilience is its form of innovation.

And together with these seemingly not-so-directly related sectors, China’s embodied AI sector comes to life.

According to industry research, China controls approximately 63% of key companies in the global supply chain for humanoid-robot components, enabling significantly lower production costs and faster time-to-market.

As mentioned, it wasn’t until the 2000s that China made significant breakthroughs in critical robotics components such as motion control, high-performance servo drives, high-precision reducers, and sensors. Domestic manufacturing of these components has gradually replaced imports, achieving precision levels comparable to international standards. A stable and well-developed domestic supply chain has supported this to come to fruition.

Share

D. Current Robotics Market Snapshot

China now runs the world’s most extensive installed base of factory robots—roughly 1.7 million units, up from 133,000 in 2013. That inventory translates to about 470 robots for every 10,000 manufacturing workers, a density that edges out Germany and Japan. Annual installations remain brisk: 290,000 new units in 2023, or just over half of the global total. Domestic production is catching up, too; Chinese firms turned out 443,000 units last year and claimed nearly a third of their home market, a steep climb from a single-digit share a decade ago. Even with softer exports and patchy consumer demand, the sector has managed to compound at roughly 20 percent a year since 2014, according to the International Federation of Robotics.

Most of those machines still cluster where speed and precision pay immediate dividends. Electronics assembly leads the pack, at about 77,000 robots installed in 2023 for fastening, testing, and packaging circuit boards and handsets. Auto plants follow with roughly 65,000 units, handling everything from welding to paint. Heavy-metal fabrication sits in third place, while plastics and chemicals are gaining ground as injection-molding cells automate picking and stacking. Even food and beverage lines are joining in, nudged by tighter hygiene rules and rising labor costs.

BUT they don’t look like humans, not yet.

These robots, mostly articulated six-axis arms, act as strong workhorses in the process, holding around half of all sales and dominating welding and general handling tasks. These industrial-use robots shuttle parts, assemble equipment with precision, and move heavy loads in warehouses on and off the assembly line across manufacturing plants.

Not until the recent 10 years did foreign names still set the technical benchmark—FANUC, ABB, KUKA (now owned by Chinese Midea), and Yaskawa hold just under 70 percent of the Chinese market. Yet local challengers are closing in. Chinese robotics company, Siasun, offers a broad catalog with Estun and Efort focused on articulated arms for carmakers. Home-grown builders started to take over some market share starting in 2010, and by 2015, when Made in China 2025 was launched, volumes of local supply went into full gear and started to take over imported makers. From 2020 onwards, with the development of connectivity, AI embodiment, and 5G networks, Chinese robot makers are leveraging rich databases and favorable policies, and have started to take over as the global innovation leaders. In short, the gap is narrowing more quickly on cost than on absolute performance, but the trajectory favors deeper domestic share.

E. Humanoid and Quadruped Robot Development

For many years, robotic arms have been programmed to help in industrial settings, mostly completing pre-programmed and repetitive tasks. Today, these robotic companies are making human-looking robots empowered by AI in hopes of actually replacing or supporting human labor in challenging tasks in labor-intensive manufacturing.

The WSJ recently wrote, similarly to what I wrote previously, that this is the convergence of technology that China has planned strategically over the decades, largely pushing forward with friendly policies and subsidies. So let’s take a look at how exactly China’s robotics sector is looking…

Embodied AI Robots

We’ve explored quite a bit about China’s humanoid robots on AI Proem, from the star company Unitree to the other start-ups offering robot maids and robot girlfriends.

The industry's growth has been fueled by a powerful combination of factors: strategic government policies, manufacturing ecosystem advantages, cross-industry technological synergies, academic-industry collaborations, and substantial investment. China has leveraged its manufacturing, battery technology, and electronics strengths to create a robust foundation for robotics innovation while simultaneously developing cutting-edge capabilities in humanoid and quadruped robots.

And I keep going back to Made in China 2025, whose goal was made explicit: Beijing wants “initial mass production of general-purpose humanoid robots” by 2025, with subsidies and AI and robotics pilot zones across the nation, spanning 11 cities from North to South, East to West.

The working-age population continues to shrink in China, raising serious concerns for economic planners. Factors include an aging population, but also a reluctance in the younger generation that is choosing to work in service industries and live in urban areas over factories, which has made the traditionally known “cheap labor” not so cheap anymore.

Today, enhancing automation in manufacturing is not just a federal-level priority, but for many, it is a firm-level priority. Engineers at UBTech are trying to train robots to sort auto parts and move containers to help with the labor shortage in mundane tasks.

Tech Diffusion

In many ways, the Chinese government has followed what Professor , author of Technology and the Rise of Great Powers: How Diffusion Shapes Economic Competition, has written: successful diffusion requires robust institutional adaptations, including education, workforce training, governance, and infrastructure that support the absorption and scaling of new technologies. Ding highlights that long-term leadership emerges from patient, strategic investments in these foundational elements rather than reactive, short-term competition or military rivalry. So in that case, we can see the robotics industrial evolution has taken decades of iteration, learning, and planning to reach the level of maturity it has reached now.

However, in the case of embodied AI, China may be adopting it at a far faster rate than the U.S.

Following Ding’s framework, China has done so to lead in robotics. Manufacturing, as a core GDP contributor, has trained millions of skilled workers. As we have also written extensively about, national-level to provincial and municipal policies have pushed forward robotics development through a series of Five-Year Plans, and the infrastructure build-up across tangentially related industries has also been prioritized.

And China is very clear on its advantages and disadvantages. Despite still lagging in the most advanced chips, it has been leveraging its manufacturing capabilities and doubling down on physical AI. From robots to embodied AI EVs, China has produced some of the most advanced and cheapest models that are being sold globally. As puts it in his comparison of China and the U.S.’s labor structure, “The shift from software to manufacturing represents more than just a sectoral rebalancing—it signals a potential return to an economy rooted in production.”

In fact, since the 2018 trade tariffs, China has been increasing its ability to be self-sufficient in many advanced tech sectors. In the case of robotics. Qiao Hong, at the Chinese Academy of Sciences and director of a laboratory of multimodal artificial intelligence systems, said domestic humanoid robots have seen continuous technological advancements in fields from component manufacturing and machine tool processing to control systems, and are now aligning with internationally advanced standards, according to China Daily.

"By combining these technologies with well-developed, independently built robotics supply chains, the country has established a solid foundation for final complete system integration," Qiao added.

"Over 95 percent of the components in our full-size humanoid robots are now domestically produced," said Zhang Dapeng, an executive at Shenzhen, Guangdong province-based humanoid robotics company Leju (Shenzhen) Robotics Co Ltd.

Market Growth and Projections

In some ways, this prioritization could also be cultural. This take could be a bit controversial, but I think there is an embedded belief by the average person that just does not get why you need to pay for something you cannot touch - in Chinese, uncles often like to scold kids by saying what they’re doing is “虚 (xu)“What does that mean? It means “empty or vain,” but it’s not that simple. The word means it’s not tangible, unreliable, or practical in an economy built up by people who made THINGS, like auto parts, textiles, and toys, to today’s EVs, home appliances, and robots. Software just seems “xu,” it is for free use, and the objective of the software should probably lead you to something that is “实 shi,” real, practical, tangible, and touchable. And what is more real than robots and machine humans taking over the heavy lifting of the laborious economy?

As of 2023-2024, China accounts for over 50% of global industrial robot installations according to China Briefing, with robot density surpassing major industrial nations like Germany and Japan. Domestic manufacturers now hold nearly half the market share, reflecting strong indigenous innovation and production capabilities. Based on Morgan Stanley’s recent Humanoid Robotics report, China is far ahead in the lead of new humanoid robot models, followed by the U.S./ Canada.

On the Big Technology Podcast, Ranjan Roy and discussed China’s humanoid robot marathon, which took place in Beijing last month. And in many ways, perhaps we don't need humanoid, human-looking robots, as it may be a bit egotistic of humans, and could not be the most efficient. There is a novelty around humanoid robots because they can fit into our existing world more easily. Rajan goes on the pod about how incredible the technology is with his DJI drones; they’re not only cheap but also just very high quality. (And thank you for the shout-out to my analysis.)

While I agree, humanoid robots currently seem a bit gimmicky. But beyond the headline hype, China's humanoid and quadruped robots sector is indeed experiencing rapid growth. The current humanoid robot market is projected to reach RMB 75 billion (USD 10.3 billion) by 2029 and potentially RMB 300 billion (USD 41.3 billion) by 2035, according to state-backed media China Daily’s report. In addition, industry forecasts suggest China is well within plan with its 2025 goals and could have 59 million humanoid robots in operation by 2035.

Subscribe now

Based on Morgan Stanley’s humanoid robotics research, 56% of the world's publicly traded humanoid robotics companies are based in China, and 61% of global robotics unveilings since 2022 have come from Chinese companies. And just looking at the first two months of 2025 alone, Chinese humanoid robotics startups raised US$276 million across 20 deals, compared to US$176 million in four transactions during the same period in 2024.

Major Players in Humanoid Robotics [see pieces I’ve written before]

• Electric Dreams, Robot Reality: China's EV-to-AI Evolution

• Physical AI’s ChatGPT “moment”: A Closer Look at Embodied AI Robots

• Robot Girlfriends, Robot Firefighters, Robot Dogs, Robot Maids and Beyond

Ok, let’s start wrapping this piece…

F. China’s Directional Policies

Hardware Advantage and Production Velocity

People like to always say China’s growth is all state-driven or directed. But it’s both yes and no. At a top-down level, there has been incredible support and encouragement to grow the AI and robotics industry. However, entrepreneurship and self-driven innovation have also played a significant role.

Policy Architecture: Beijing’s Roadmap for Embodied AI

The state has scaffolded that hardware momentum with an unusually dense policy support, and undeniably, these have played the most significant accelerators:

• 14th Five-Year Robotics Plan (2021-25): The Ministry of Industry and Information Technology’s five-year blueprint calls for breakthroughs in “brain, cerebellum, and limb” technologies and for the formation of three to five globally competitive robot clusters.

• Made in China 2025: Robotics sits among ten “strategic” sectors where import reliance must fall and domestic market share must rise; recent reviews show the programme has already narrowed many hardware gap.

• “Robotics+” Application Action Plan (Jan 2023) — Co-signed by MIIT and 17 other agencies, the plan bankrolls deployments in factories, warehouses, hospitals, and consumer services, ensuring an on-shore demand pipeline for early-generation humanoids.

• Government Work Report (March 2025): The latest major push came from the top down. China’s premier annual policy address listed embodied intelligence alongside quantum tech and 6G as future industries eligible for increased funding and pilot zones, locking humanoids into the national innovation budget, according to NPC Observer.
  

  Share

China’s humanoid and quadruped builders have also been battling it out domestically in a fierce competition of technology and price competitiveness. As mentioned, the price edge means that finished machines typically sell for 30-50 percent less than comparable Western models, a gap rooted in shorter supply chains and aggressive component integration, according to ITIF. Lower sticker prices, in turn, widen the customer pool and justify larger production runs. Firms such as Unitree, Fourier, and Agibot already plan thousands of units a year. Just Unitree alone expects domestic makers to ship more than 1,000 humanoids in 2025, while Agibot targets 5,000 units; these numbers still seem aspirational in the United States or Europe.

Together, these layers push capital toward both high-volume factories and component R&D, while the local government grant system offers subsidies to early adopters on everything from reducers to cloud-robotics subscriptions. The result is a feedback loop: cheaper parts enable bigger production runs, which generate more field data, improve AI policies, and still attract more customers. And you can say, policy may have proven as decisive as engineering in China’s race to mainstream humanoid robots.

G. Conclusion: Progress with Imminent Challenges

The overarching geopolitical tension between China and the U.S. makes the advancement of these technologies more scrutinized. Although many are for commercial use, the advancement of robotics, especially embodied AI robotics, has now been labeled a national security issue.

Limits are at the technical and potential economic levels. AI chips, for one, are one of the biggest hurdles the industry faces in China. Precision gearboxes, high‑end sensors, and purpose‑built AI chips remain bottlenecks, leaving builders reliant on imports for the most exacting components. At home, regulators are only beginning to draft safety and performance standards for free‑roaming humanoids, a process that must keep pace with rapid deployment. Economically, the country will have to shepherd displaced workers into new roles as automation scales, a social‑policy challenge as technical as the robots.

Finally, each new cohort of machines adds to demand for electricity, rare‑earth magnets, and battery metals—resources already under pressure from the EV and renewables booms. This forces planners to balance industrial ambition with energy and materials security.

China's robotics industry has grown from early experimental efforts in the 1970s to become the world's largest and most dynamic robotics market. And as China continues to advance its robotics capabilities, the industry is poised for further growth and global influence. The combination of technological innovation, manufacturing prowess, policy support, and entrepreneurial dynamism suggests that China will (for now) remain at the forefront of the worldwide robotics revolution for the foreseeable future.

But again, real-scale usage of humanoids across workplaces? Probably not for another ~3 years at least, I’d say.

For a while now, and I have been working on an analysis of China’s prioritization of “实体经济” the so-called “real economy” [which we’ll aim to get it out there in the next few weeks] and how that has laid the foundation of its direction of AI building. We touched on it briefly as one of the factors explaining why China’s AI monetization strategy has been heavily skewed towards the consumer application end vs. the enterprise strategy the U.S. AI labs are adopting, as China’s knowledge worker base makes up such a small percentage of its workforce that is mostly in industrial roles driving the growth we’ve witnessed over the last few decades. This all relates to our story - EVs’ integration of AI & venture into robotics.

To understand how Chinese EVs are becoming AI pioneers, we must first examine the ecosystem that enabled their rise to global prominence.

Subscribe now

China's EV Dominance: A Convergence of Strategic Industries

As the age of embodied AI dawns, China's bedrock of manufacturing prowess uniquely positions it at the forefront. But before we understand how Chinese electric vehicles (EVs) are pivoting to AI, let’s understand how Chinese EVs even became global industry leaders. A concept I really appreciated that Princeton post-doctoral researcher Kyle Chan wrote in his newsletter astutely observed that numerous crucial industries in China "grew up" in tandem. This wasn’t a matter of chance but a deliberate national planning effort, meticulously fostering a network of interconnected sectors. It was a conscious national planning effort to push forward a series of connecting sectors. These sectors include lithium batteries (i.e., CATL, CALB), Smartphones (i.e., Xiaomi, Huawei, Oppo), traditional OEMs (i.e., BYD, SAIC, BAIC), lidar and sensors (i.e., Hesai, Robosense), industrial robotics (i.e., Inovance, Siasun), and even renewable energy solutions (i.e., Longi, Jinko Solar, Sungrow).

In many ways, all of these were propelled forward through the country’s Five Year Plans, one five-year at a time, and now have converged at this point in time to be ideally suited to enable the next generation of embodied AI EVs.

This tech-industrial convergence, a theme I’ve previously explored in my Physical AI primer, reveals that the underlying technologies of EVs, robotics, and AI are not novel in isolation. The actual inflection point lies in their synergistic convergence, where Chinese EV market leaders find themselves remarkably well-positioned, possessing accessible technologies that are now converging and coming together along the supply chain.

In a December 2024 report, the Financial Times reported that EVs were to outsell traditional cars in China for the first time in 2025, putting China’s EV transition adoption miles ahead of major Western markets. The same report said that Beijing’s official target for domestic EV sales (including battery and plug-in hybrids) is growing by about 20% year over year to more than 12 million cars in 2025.

Robert Liew, director of APAC renewables research at Wood Mackenzie, said that China’s EV milestone signals its success in its domestic technological development and its ability to secure global supply chains for critical resources used in EV batteries. Adding that “they want to electrify everything…no other country comes close to China,” Liew told the FT.

I spoke to Cameron Johnson, Partner at TidalWave Solutions, a Shanghai-based supply chain management consultancy. Cameron is a regular on Bloomberg TV, commenting on China’s economy, industrial growth, and supply chain-related issues. You might recognize him from there ;p. He’s provided some excellent insights into how to view China’s AI advantages through its manufacturing sector.

The Human Element: China's Technical Talent Advantage

While manufacturing capabilities and supply chains are crucial, China's competitive edge in AI and EVs is equally rooted in its human capital development.

And I think the talent aspect has often been overlooked, as I’ve written before that Zhejiang University educated three out of six of the AI dragons in Hangzhou.

Beyond that, “China graduates roughly 1.5-1.6 million engineers per year, vs. 100K+ in the US,” said Johnson; just sheer number-wise, it’s being underappreciated. Adding that, “People miss the ecosystem that is powering these industries in China. The integration of the supply chain ecosystem fueling these (and other) areas is not considered when people look at what is happening in China.”

He’s not the only one saying that. In ’s writing, he also pointed out that industry leaders in the U.S. actually know about this.

Hua wrote: “Pavel Durov, the founder of Telegram, attributed China’s rapid AI advancements to its highly competitive education system, pointing out China’s consistent excellence in global math and programming competitions.”

MarcoPolo, a think tank part of the Chicago-based Paulson Institute, found that "researchers originally from China now make up 38 percent of the top A.I. researchers working in the United States.” I mean not to stereotype, but even visually, we’ve seen many East Asian/ Chinese-heritage engineers in launch demos from all the U.S. AI labs; of course, many are second—or third-generation immigrants like myself, but many actually only migrated to the U.S. in recent years.

And we’re not just talking about programming engineers. About ten years ago, the rise of DJI, SenseTime, Horizon Robotics, and similar firms trained a generation of hardware-software engineers. These engineers are now out in the world creating something new, leveraging what they learned at these leading drone and industrial robotics companies and making them more sexy—humanoid robots.

Even Jensen Huang, the OG himself, said that China has a strong STEM education and trained robotics and mechatronics capabilities during his HKUST fireside chat last November.

Global Headwinds: Navigating Market Access Challenges

Despite China's technological achievements in the EV sector, geopolitical tensions have created significant market access challenges that shape how these companies strategize their global expansion.

China's EV battery manufacturers, including CATL and BYD, produced more than half of the world’s EV batteries in the first half of 2024. But that doesn’t mean they’ve had it easy. Despite global dependence on China's EV battery sector, Chinese electric vehicles face increasing restrictions across Western markets due to competitive protection concerns and increasingly geopolitically charged narratives.

The United States, European Union, and Canada have each implemented significant restrictions on Chinese EVs, though their approaches and severity levels differ. As the China-US trade war continues to normalize as a backdrop to international trade and competition and concerns about advanced technology security intensify, countries like Canada have cited issues regarding labor conditions in EV manufacturing. China has responded by filing complaints with the World Trade Organization, viewing these tariffs and restrictions as protectionism while directing Chinese OEMs to reduce their investments in European counterparts.

Meg Rithmire, a Harvard Business School professor who researches China’s political economy, told Rest of World. Adding that many recent incidents have shown tightened party control over Chinese society, state investments in strategic industries, and Beijing’s more aggressive foreign policy have all fueled a deeper suspicion in the U.S. (and the West as a whole)

“There’s just this blurred boundary between firms and the state,” said Rithmire. (from Rest of World’s deep dive on Chinese EV battery firm Gotion’s failed efforts to build a factory in the U.S.)

Why are EV Firms Pivoting to AI?

With a strong manufacturing base and technical talent pool, Chinese EV makers are uniquely positioned to integrate AI capabilities, creating new value propositions beyond traditional mobility. And AI seems like the most natural next thing to tap into.

On Feb. 10, BYD's shares surged to a record high in Hong Kong after announcing that it would release extremely cheap models with autonomous driving capabilities. In addition, the company said that it will integrate DeepSeek into its system to enhance the AI assistant experience, marking a strategic shift toward smart driving systems.

“Advanced smart driving will become a standard safety feature similar to seatbelts and airbags, BYD’s founder and chairman Wang Chuanfu said at a China-focused launch event livestreamed late Monday,” according to CNBC’s report. All for the cost of under RMB 70,000, that is under USD 10,000.

But looking into the future, it’s not just autonomous cars. I mean, Waymo has partnerships with Geely’s Zeekr and all, and the future of mobility may soon be driverless, but that’s not the focus of this piece today…

Future of EVs = Electrification + Intelligence

“The unveiling of Deepseek was an earthquake on a global scale,” said Johnson. We’ve seen DeepSeek being integrated swiftly across industries in China. The integration of DeepSeek into China’s EV space has been swift.

Many key industry leaders, from BYD to Li Auto, share this vision - AI can be empowering the cockpit. Notably, Li Auto’s CEO, Li Xiang (yes, he named the Chinese brand after himself), believes that the "electrification" of electric vehicles is only the first half of the game. The true second half is "intelligence." However, in a recent interview, he said that the intelligence here is not traditional software intelligence but a complete evolution empowered by AI.

Adding that, Li Auto’s ambition is to establish a leading artificial intelligence enterprise, where automobiles serve as the means, and intelligence is the ultimate goal."The vehicle" serves as a vessel for Li Xiang's vision of achieving intelligence, or rather, it is a profitable product.

Government Fueling the Embodied AI Revolution

Government incentives again have played a huge role in bolstering this industry. Like many other sectors under “the real economy,” municipal and federal governments have helped elevate them. For example, in Shenzhen, the local government has just implemented policies to boost AI and robotics development. By 2027, it aims to host more than 10 companies valued at over 10 billion yuan ($1.3 billion) and over 20 companies with annual revenues exceeding 1 billion yuan. And according to Nikkei’s reporting, that’s only part of a string of incentives. Shanghai aims to cultivate a 1 trillion yuan industry by the end of this year. At the same time, Beijing has announced plans to establish a 10 billion yuan robotics fund and create a humanoid innovation center. Provinces such as Zhejiang, Shandong, Chongqing, and Anhui have all rolled out targeted robotics action plans.

According to the government's official press release, “AI Plus” was introduced in the most recent Two Sessions meetings. This initiative aims to propel China into an AI "super market” through clearer regulatory support and control of AI misuse, educational efforts to address talent shortage issues, and integration of AI across industries, such as EV, to help critical industries.

According to ’s most recent analysis. China’s robotics is also going full-local, and we’ve discussed how that cost advantage is enormous in the Unitree piece. Semianlaysis wrote that the localization effort is well underway.

“Local firms are taking over the world’s largest market, approaching a 50% market share, compared to just 30% in 2020…The rise of Unitree exemplifies this shift: the only viable humanoid robot on the market, the Unitree G1, is now entirely decoupled from American components…Today, building an identical robot arm (modeled after the Universal Robots UR5e) in the US is ~2.2x more expensive than in China. Under the hood, the situation is even more alarming.”

It’s everything from policies to cost advantages to know-how, underappreciated talent, and government incentives that are really helping Chinese EVs pivot to AI so quickly and smoothly.

Capability Overlap: Autonomous Driving Technology and Robots

To be honest, the demand for humanoid robots isn’t new to China. There have been years of need for manufacturing automation as fewer blue-collar workers want to work manual labor jobs and an increasing issue with labor shortage due to an aging population. China has also had an active national push to dominate humanoid robotics by 2027, and we’ve seen existing supply chains and experience with advanced technology.

In a Nikkei article, Paul Gong, head of China autos research at UBS Investment Bank, described Musk as a sort of "spiritual leader" for the humanoid industry, a role he also played in EVs. Tesla's full self-driving technology, he said, "pushed Chinese automakers to continuously pursue smarter vehicles and autonomous driving, and the introduction of the humanoid robot concept has also led numerous Chinese companies to explore this direction."

We’re seeing that EV players' mainstream interest in humanoid robots took off after Tesla released its Optimus robot in 2022. Multiple Chinese companies already announced last year that they had started to deliver humanoid robots to EV makers. Chinese EV manufacturers are leveraging their technological expertise in sensors, batteries, and machinery to develop these humanoid robots - all making a lot of sense when you spell it out like this.

In 2024, China's humanoid robot market reached 2.76 billion yuan, with forecasts projecting growth to 75 billion yuan by 2029, representing 32.7% of the global market.

To note what a few major players have done: BYD has initiated a humanoid robot project, codenamed "Yao Shun Yu. " GAC Group unveiled "GoMate," its third-generation intelligent humanoid robot, in December 2024. XPeng unveiled its humanoid robot called "Iron" in November 2024. And NIO is recruiting heavily in the space and investing heavily in R&D.

But for any of these products to go to market safely, with proper regulation and consumer trust… well, that might take some more time… today, we end here. Do leave a comment or DM me if you have any feedback.

The once-promising, vibrant USD investment sector (public and private) in China’s tech industry was further stalled when the U.S. government restricted the investment of U.S. citizens and firms across all asset classes in Chinese Semiconductors and microelectronics, Quantum computing and quantum information technologies, and Artificial Intelligence (AI).

However, despite being unable to invest directly in companies like DeepSeek or Unitree, some investors have thought of a clever way to invest in ADRs as a safe way to gain exposure to them. That is, assuming that Alibaba (BABA) or Baidu (BIDU) (and other similar companies) in the world would at least be able to access these technologies or benefit from their developments.

Chinese tech, AI, and robotics firms are suddenly receiving all the global attention again. Thus, today, we introduce you to Unitree Robotics, a leading robotics innovator.

This post was first published on as a guest post on Jan 22, 2025. Check out ’s newsletter, which covers everything about AI under the sun.

At CES 2025, it was all about AI integration into hardware, from wearables to AI-empowered furniture to robots! And not the non-sexy machinery robots that warehouses have all adopted—the fun ones, the ones we see in sci-fi movies.

There were 3,800 exhibitors at CES 2025, 1500 of whom were from China. If people said AI integration was still nascent in 2024, 2025 showed concepts becoming reality. Intelligent robotics completely dominated the showrooms, showcasing the rapid advancements in AI integration, from robotic dogs to AMRs to cleaning, agricultural, and compassion robots. Physical AI is having a ChatGPT moment.

That brings me to today's protagonist: the Chinese robotics firm Unitree and its robot dogs, which completely stole the show and have been the subject of much hype and craze on social media.

[On the All-In Podcast, the four American business/tech executives/ investors, Chamath Palihapitiya, Jason Calacanis, Gavin Baker, and David Friedberg (minus David Sacks), discussed the impressiveness of the Unitree robots and how some of them initially even thought the videos were AI-generated during their 2025 Predictions episode.]

Alas, the conclusion is that 2025 will be a big year for AI-embodied robots.

Screenshot of Unitree’s demo video

Who is Unitree Robotics 宇树科技?

Unitree Robotics, a Chinese company founded in 2016 by Wang Xingxing, has rapidly become a prominent player in the quadruped robot market. Quadruped robots are autonomous or semi-autonomous robots that mimic the four-legged locomotion of animals, in this case, robot dogs.

Amid the hype of “embodied AI,” it's receiving much attention from investors globally, including tech executives on the All In Podcast last week. The hosts were shocked by its capabilities, with David Friedberg claiming it’ll be “the year of the robots.”

Now, the company has often been compared to Boston Dynamics, but offering ⅕ of the price for similar robots (we’re seeing a trend here with Chinese innovation; see my piece on DeeSeek). The affordability and accessibility make its robots appealing to a broader range of consumers and industries, for which we’ll go into detail about product capabilities and controversies in another piece. Today, this article will focus on the founder’s story and the firm's founding background and vision. In a follow-up Part 2 to UniTree, we’ll dive deeper into humanoid and quadruped robotics use cases, potential safety concerns and implications, and the bigger picture of China’s robotics sector.

Share

Who is the Founder, Wang Xing Xing?

Wang Xingxing, the founder, CEO, and CTO of Unitree Robotics, was born in 1990 in Ningbo, Zhejiang, an industrial city south of Shanghai. He earned his bachelor's degree from Zhejiang Sci-Tech University. He then obtained a master's degree in Mechanical Engineering from Shanghai University. The story of Unitree Robotics began in 2013 when Wang, a graduate student at Shanghai University, embarked on a project to develop a quadruped robot powered by low-cost electric motors. His vision was to create a high-performance robot that was more affordable and agile than the hydraulically powered robots prevalent at the time. This project culminated in the creation of one of their first iteration, the XDog, in 2016, which was his master’s thesis that later on became a prototype that garnered significant attention online and attracted investors.

Source Sina Finance: Early picture of Wang with his first robot- Xdog.

He once said it’s all about timing, and sometimes things cannot wait. So, after receiving some media and investor attention and inbound inquiries, he deliberated for a few months and decided to capture the opportunities presented to him during China’s rampant VC heydays. Wang, who was only a few months into his DJI job, pulled the trigger, left his job, and officially launched Unitree Robotics in May 2016 with the help of an angel investor.

In an interview, he said with some shyness that the early stages of entrepreneurship were genuinely memorable, filled with bitter-sweet memories. He recalled that when he first started, he bummed on his friend’s couch for a month, and because the sofa wasn’t very long, Wang woke up with sore legs every day from dangling halfway off the sofa. He smirked but said, “Look, it’s not like when moguls start a new company, and they can just wave their hands, and a lineup of impressive talents come following; our business started very grassroots.”

Looking back at how he got to where he is now, Wang candidly said he originally wanted to go to Zhejiang University (one of China’s top three universities). His grades were good, but his English scores did not reach the minimum requirement. Hence, he chose his second-best option, the Shanghai University Robotics program. He said, looking back, “塞翁失馬焉知非福” which means “it was a loss but also a blessing in disguise” because if he went to Zhejiang University, he might have never started his own company.

Wang said he grew up tweaking around electronics and was always able to build out simple robots cheaply as he did not have much financial means. His natural nerdiness was more evident in his earlier-day interviews where instead of envisioning any grand humanoid robotics use case or AI-empowered future, he said his vision was simply that maybe one day, everyone can have a robot companion at home.

Unitree quickly gained attention from potential buyers within the 2B and 2C markets. Wang eventually set up shop in Hangzhou (home to Alibaba, Ant Financial, and Geely, to name a few), a talent-dense and manufacturing-advantaged city that is not as well known in the West. Its proximity to bustling Shanghai (the capital of entertainment, money, art, education, and international access) has also made it a top choice for many to settle there.

In 2017, Unitree released its first commercial quadruped robot, Laikago. He said he named the first product Laikago because it was the name of the first Soviet Dog to enter space and orbit Earth. Like Laika, he hopes his products are at the frontier of science and technology (although sadly, Laida died only hours into the flight).

In 2021, KrAsia (the English arm of China’s leading tech media 36Kr) wrote about Unitree in the context of how it competes with Boston Dynamics, and that is also when I first heard of the company.

Now, in multiple interviews, Wang was asked how his company stacks up to Boston Dynamics, and he always says they may look the same, but the core is different. Once he answered, Unitree allows users to take them apart and rebuild the robot or program it as they wish. Boston Dynamics is like Apple iPhones; you get what you buy and can't reprogram or set the system, and Unitree is simply the most cost-effective product out there.

Here, I want to provide some context. People often say DJI and SenseTime groomed a generation of top-notch hardware-software engineers in China. Many that have left the firms have started businesses independently, mostly setting up shop across the Greater Bay Area.

The Greater Bay Area (GBA) connects nine major cities in Guangdong province and the unique administrative regions of Hong Kong and Macau. And it is home to many of China’s biggest tech companies, including Tencent, Huawei, ZTE, BYD, DJI, and SenseTime, to name a few. It has the advantage of having access to top developers and engineers in Shenzhen and access to manufacturing capabilities in nearby small cities in Guangdong province.

Even Jensen Huang said during his HKUST fireside chat that he believes the GBA (China to some degree) holds a significant advantage in AI development due to its access to top talent and capital (from companies like DJI, Huawei, and Tencent. Huang emphasized that this area excels in mechatronics, which combines mechanical technology and electronics – AI robotics that can comprehend physical electronics and reasoning intelligence.

Subscribe now

Chinese Robots Taking over the World

Unitree is now a global leader but once had no international ambitions.

In one of Wang’s earlier interviews, he was asked by a Chinese tech publication whether he would be looking to compete with Boston Robotics and sell his products to the West. Wang Xingxing smiled and said, "I won't have this difficulty because my English is poor, and it might be challenging for me to go abroad so that this situation won't happen, and I don't need to consider this issue. But if it were before starting my business, I might have gone, but I haven't thought about it now."

That attitude has obviously changed over the years.

According to data from the Gaogong Robotics Industry Research Institute (GGII), Unitree Robotics accounted for 69.75% of global sales and 40.65% of the market share in the quadruped robot industry in 2023, solidifying its position as the sector's leading enterprise.

To understand how Unitree fits into the bigger picture of China’s robotics space, we can see that in 2023, the Ministry of Industry and Information Technology laid out a blueprint encouraging talent, capital, and government resources to cultivate robotics development. The paper outlines the state’s ambitious goals to leverage its strength in manufacturing and integrate artificial intelligence into its robot, as well as plans to establish a reliable industry supply chain system to support its manufacturing. Bloomberg wrote a short write-up about it here in 2023.

A Global Times article, a news outlet backed by the state propaganda department, covered the World Robot Conference in Beijing and cited Liu Cong, the Vice President of iFlytek. He said he expects humanoid robots to enter various industries in the next 3-5 years and households in the next 5-10 years. This is all part of the state-encouraged direction to innovate in software and hardware integration.

We can see how China has become a leader in this space. Simply looking at the leading robotics firms in the chart above, 90% are Chinese, and more than half of the exhibitors at CES 2025 are Chinese.

Unitree’s Funding History

Wang said in its early stages, he didn't have any money, but he still counts himself lucky. There was a lot of encouragement from his peers, and the overall environment was very encouraging for him to pursue entrepreneurship. And because of the lack of significant funding, beyond his dedication to research and creating a viable product, he was always sure to commercialize his robots from the start.

A VC investor I know met with the company leaders in 2019, but the investors did not see their scalable use case back then, and many VCs back then focused on industrial robotics (such as Hai Robotics at the time). Before AI integration, robot use cases seemed better positioned to replace hard and difficult laborious tasks. However, as AI and robotics merged, AI-empowered robotics opened investors' eyes to a completely different potential TAM.

As Unitree survived the COVID pandemic and the USD investment pull-out of China, it has become one of the only (few) robotics companies that survived the long winter and can now manufacture in mass.

In 2016, at the start of Wang’s entrepreneurial journey, Unitree Robotics secured a 2 million RMB angel investment, as mentioned above, that’s just under 300,000 USD.

Fast forward, Unitree has successfully secured its Series C funding of US$150 million from investors, including HongShan (Sequoia’s China spin-off), CITIC Securities (China’s leading financial conglomerate), and Jingguorui Equity Investment, among others, in September 2024, according to CB Insights, a market intelligence platform.

Since its inception in 2016, Unitree Robotics has undergone nine funding rounds. The latest round raised 1 billion yuan, valuing the company at over 6 billion yuan.

The most recent B2 funding round, completed in February 2024, raised RMB 1 billion and valued the company at over RMB 6 billion (about USD 820 million). Investors included Meituan, Jinshi Investment, and Source Code Capital, among other prominent funds. Existing shareholders such as Shenzhen Capital Group, China Internet Investment Fund, Rongyi Investment, Dunhong Capital, and Midascore followed with additional investments.

Prominent investor comments (translations from 36kr):

• The lead investor in the most recent fundraising round, Chen Pingjin, Deputy General Manager of Jinshi Investment, said, "[We are] optimistic about the application prospects of quadruped/biped robots in consumer entertainment, industrial applications, educational research, and other fields. We fully recognize Unitree Robotics' leading position, with its capabilities in independent R&D of software and hardware, cost control, and platform construction. We are confident in the company's growth potential." During an interview with 36kr.

• Chen Runze, Executive Director of Source Code Capital, another investor in this round, commented: "Robotics and automation are critical endowments of China's technological innovation. In recent years, driven by advancements in AI technology, robots have made significant progress in perception, motion control, autonomous navigation, and operational capabilities, creating opportunities for deeper market penetration. Through our long-term tracking of the industry, we have observed Unitree Robotics' steadfast dedication to legged robots, driving the commercialization and mass production of next-generation robots and achieving a leading position globally. At the same time, the company continues to break new ground in the technology stack of general-purpose robots. With its prudent commercialization strategy, strong R&D capabilities, and the team's ambitious vision, we believe Unitree Robotics has the potential to become a world-class leader in the robotics industry."

• As a two-time investor in Unitree Robotics, Liu Hongchun, Founding Partner of Rongyi Investment, remarked: "Humanoid robots are approaching a critical juncture in the evolution toward general-purpose robotics. Unitree Robotics has once again demonstrated its pioneering capabilities in developing general-purpose humanoid robot technology. The company's self-developed joint motors are smaller, lighter, and more potent than similar humanoid robots. As investors, we are highly optimistic about Unitree Robotics' technology, products, and ability to integrate with real-world business scenarios. We look forward to the advent of the 'embodied intelligence' era led by general-purpose humanoid robots."

Subscribe now

Rise to Fame

Wang never seemed to shy away from the public despite his seemingly demure presence. The company has actively done some high-profile PR.

One of its most flashy peacocking efforts was in 2022. Unitree showcased a choreographed dance by its robots as part of the Beijing Olympics opening ceremony. This was an opportunity, an honor, and a clear approval/ acknowledgment by the authorities.

The characters above are demonstrated by the robots' spell “Winter Olympics.”

In just the past year, the firm seemed to have engaged in a series of PR efforts to promote its leading position in integrating AI within its robots, from interviews with Technode, a leading China-focused tech digital publication, to Global Times, China’s state-backed media.

Specifically, in the interview with Technode, Unitree’s cofounder, Chen Li, said amid all the AI breakthroughs, Unitree “firmly believe(s) that once general artificial intelligence and general humanoid robots are combined, robots can truly enter daily life in the future.” They aim to “integrate these latest artificial intelligence technologies with robotics to take robot functionality to the next level more effectively.”

He added that robot technology has been at the forefront of research and technological innovation in Japan and the U.S. for over two to three decades. However, we may see more consumer adoption as AI is integrated into robotics technology.

Looking at how the company has evolved over the eight years since its founding, he told TechNote that Unitree was initially a relatively traditional robotics company. Still, it is gradually transforming into a company that integrates “AI+ general-purpose robots.”

Over the years, the team has conducted extensive strategic brainstorming to envision scenario use cases, from household use to firefighting, entertainment purposes, and more. He thinks that general-purpose household robots can assist homeowners in various tasks such as washing dishes, cleaning floors, making beds, and cooking.

However, single-function robots are insufficient to accomplish these diverse tasks. To achieve such a vision, he added that robots must be empowered with general artificial intelligence, and this empowerment will need to be achieved through end-to-end approaches, cloud-based methods, or training methods.

[Friend said she saw a person walking one of these robot-dogs in the neighborhood park 2 years ago in Shanghai, waawhatt!]

Check out Part 2 for a more in-depth use-case analysis.

Part 1 Founding Story Parting Comments

With the executive order restricting U.S. firms from investing in sensitive industries, advanced technology such as humanoid robotics and AI is mostly only backed by domestic capital in China. This starkly contrasts the consumer internet era when U.S. investors held some of the most prominent positions in Chinese mobile apps.

We’ll continue to see that talents from the two super nations will innovate in parallel; we'll need to wait and see whether further restrictions will be put on the import/export of AI-powered hardware. But even just Monday this week, we saw that the Biden administration pushed out probably one of its last significant policies as it tries to ensure allies comply with American AI standards, govern AI’s spread across tier-ed nations, and further curb China’s access to advanced technology.

If you're interested, the CFR and Carnegie Endowment for International Peace provide more detailed analyses of the policy’s specifics.

Nonetheless, we can expect more robotic technological breakthroughs in 2025. So much innovation has been happening away from the general public’s eyes, and that is now finally being revealed. We will see more application-end innovation as AI evolves and matures, including software-hardware integration.

In October 2024, during a video call at a tech conference in Riyadh​​, Elon Musk predicted that by 2040, there would be at least 10 billion humanoid robots priced between $20,000 and $25,000 through a video call at a tech conference in Riyadh. We are still far from that, but we are one step closer.

Link to Part 2: Robot Girlfriends, Robot Firefighters, Robot Dogs, Robot Maids and Beyond

Recent top 5 reads from Grace:

1. Everybody Losing Sleep Over DeepSeek: Industry Implications to LLMs and AI Infrastructure

2. China’s ChatGPT Doubao shows that inference compute demand COULD actually go up by 1 billion times.

3. The AI arms race is far from over: chips are only half the game, and infrastructure is the other

4. Behind the AI Arms Race: U.S. vs. China Cloud Computing

5. DeepSeek V3 puts China AI on the global map: consumer use and capital expenditure implications

This has been a learning journey, both about the topics I cover and about running a successful newsletter. As I continue deepening my research, I aim to improve reader engagement. If you have the time, please DM me through the platform to provide feedback.

It also means a lot and helps my content reach more people if you like and share the articles you enjoy. This will improve the feed recommendation algorithm and help grow. Thank you!

Subscribe now

Following my guest post for , written in collaboration with on Physical AI, humanoid robots, and a deep dive into Unitree, this is Part 2. For Part 1, see here: Introducing Unitree, China’s leading AI-embodied Robotics Company.

Many knowledgeable analysts and writers are covering U.S. robotics development on Substack, from Tesla’s Optimus, Figure AI, Boston Dynamics, Agility, and many more. Today, I want to do something that not as many people are doing on Substack: dive deeper into the Chinese robotics landscape simply because - LOOK at the chart below.

But first, let’s quickly wrap up the Unitree section. As promised, we’ll look at the robots’ specs and use cases.

Unitree Part 2

Unitree Robotics offers a diverse range of quadruped robots designed for various applications, and they’re all about being “agile and affordable,” they categorized the bots into several series:

• Go Series (the robot dogs): Designed for consumers and researchers, featuring advanced AI capabilities, a compact design, and high speed.

• B Series (for business use): Industrial-grade quadruped robots with high payload capacity and enhanced environmental tolerance for demanding tasks.

• Humanoid Robots (the human-looking ones): Advanced humanoid robots designed for research, development, and potential applications in various fields.

• LiDAR Sensors: High-performance LiDAR sensors for enhanced perception and navigation in robotics and autonomous systems.

This is the Go2 Robot Dogs that took over the internet

What is shocking is that the Go2 robot without a controller sells for only $1600 on its website—less than the cost of some gaming consoles. According to a Chinese VC investor, most of the robots are currently sold to universities and research labs internationally, and they’re mostly used for program training and research. Though cheap, people have not yet found a significant mass market need.

In terms of applications, think of it like this: consumer robots focus on home use—entertainment and companionship—while the B2 robots are often designed for demanding tasks in sectors like energy and emergency services, handling inspections, and rescue operations.

The company has said that there are many different potential use cases (with the help of Perplexity):

• Inspection and Surveillance: Robots like the B2 can perform inspections in industrial settings and are equipped with thermal and gas detection modules to monitor hazardous areas efficiently.

• Logistics and Delivery: The B2 robots excel at transporting goods in warehouses and outdoor environments. Thanks to its robust payload capacities, it can handle heavy loads over extended periods.

• Research and Education: The Go1 EDU and humanoid robots serve as educational tools in robotics and AI, providing hands-on learning experiences through advanced sensors and programming capabilities.

• Disaster Relief: Unitree robots are designed to navigate hazardous terrains, aid search and rescue operations by carrying supplies, and assist responders.

• Defense Applications: The Go1 has been utilized in military drills, showcasing its potential for tactical training and operations.

The key distinctions between Unitree's consumer-grade and industry-grade robots are pronounced based on the co-founder’s analysis, and they’re mainly in these areas:

• Performance: The B2 industrial robot boasts a payload capacity of 50 kg, far exceeding the Go2 AIR's 5-8 kg limit.

• Protection Level: The B2 achieves an IP68 rating, ensuring it is waterproof and dustproof, while the Go2 lacks such protection.

• Mobility: The B2 can navigate obstacles up to 1.2 meters high, compared to the Go2's limit of just 16 cm.

• Battery Life: The Go2 operates for about two hours on a single charge; the B2 lasts over five hours.

They’re probably one of the best-selling robots right now. Still, we’ve not seen mass adoption in our day-to-day lives yet due to technical, regulatory, and ethical challenges, which we’ll discuss further below. As of now, Unitree Robotics is probably the closest one to be able to reach us by mass. The CEO, Wang Xing Xing, has vowed at the 2024 CES that he plans to keep humanoid robots priced below $90,000 to enhance accessibility.

Ok, that’s it for Unitree. To learn more about the company’s founding story and Wang Xing Xing, check out Part 1. Let’s zoom out a bit.1

Subscribe now

The Convergence of AI and Robotics

We’re at a tipping point for autonomous humanoid robots, where the convergence of GenAI technology and machinery is happening. Here, I wrote a basic primer on Physical AI and what makes humanoid robotics special.

In China, specifically, robots are taking over the world quicker than people expected. “China does not yet appear to be leading in robotic innovation, but … it is likely only a matter of time before Chinese robotics companies catch up to the leading edge,” the Information Technology and Innovation Foundation (ITIF) said in a report released in March 2024, based on SCMP’s report.

Leave a comment

Earlier this week, Nvidia's Founder and CEO, Jensen Huang, spent five days in major Asian cities: Shenzhen, Taichung, and Beijing. He engaged in low-key discussions with partner companies on cutting-edge technology topics such as liquid cooling, silicon photonics, embodied intelligence, and robotics.

With Huang’s recent firm conviction that Physical AI is the next era of AI (expressed at CES and quite literally everywhere he speaks), we can see that he planned strategic meetings with the founders of robotics innovators coming out of China. It’s consistent with his messaging and his interests. He said at his HKUST fireside chat late last year that China has an advantage in robotics, given its know-how in mechatronics. With the rise of DJI and SenseTime, a generation of strong hardware/ software integration engineers were trained in China over the last decade.

China’s Major Robotics Players

Judging from the leaked seating arrangement on WeChat of the appreciation banquet in Beijing, numerous luminaries from the humanoid robotics sphere were invited to attend. Among them were Wang He, the founder of Galbot Robotics; Wang Xingxing, CEO and CTO of UniTree Robotics; Nie Xiangru, co-founder and COO of Paxini Tech (TORA-ONE); and Professor He Bin from Tongji University.

A screenshot of the founder, He Wang, of Galbot’s WeChat moments went viral, referring to Huang as the “godfather.” Wang wrote: “The first thing that Godfather said to me was ‘I saw Galbot. That’s really cool.” He added that “partnership and recognition are mutual and are based on shared values” while sharing a picture of the seating arrangement where he sat next to Jensen during the dinner.

Other notable founders at the banquet obviously include our recently covered protagonist, Unitree Robotics’ Wang Xing Xing.

This wasn't the only big AI gathering this week in Beijing. DeepSeek’s founder, Liang Wen Feng, met with Chinese Premier Li Qiang at a meeting, where the senior politician met with various business leaders to get their feedback. This event was widely covered by China’s national state broadcaster, CCTV.

In addition, at the World Economic Forum in Davos, Switzerland, Chinese Vice Premier Ding Xuexiang said yesterday that China is developing AI to accelerate economic transformation, achieve Chinese modernization, and enhance the quality of life for its more than 1.4 billion population. He added that the country has implemented robust regulatory frameworks and mechanisms to ensure the responsible use and management of AI technologies.

You can say, it’s an all-hands-on-deck and let’s build some AI/ robots kind of moment for Beijing.

Subscribe now

Real-Life Use Cases and Risks

Looking at real-life use cases, it’s crazy to imagine that the available robots can be used in so many ways, basically in any way a human could function. wrote about LAFD using robots to help with the recent tragic fire engulfing Los Angeles. Maybe a more advanced humanoid version could even help rescue people from fire one day.

In the Galbot video, you can see the humanoid robot has four main application use cases: home for cleaning, retail for restocking shelves, manufacturing such as lifting and transporting goods, and helping out at the pharmacy. Here, it highlights the home scenario.

The Unitree humanoid robots can be folded up for transportation when shut off. When turned on, it can be your running companion, help you open a Coca-Cola bottle, flip a pancake in a pan, and fight you with a stick. This is absolutely wild.

And in this case, China will host the world’s first human-robot marathon in the Daxing district in Beijing this April. About 12,000 humans and robots from more than 20 companies will partake in the 21 km race.

Hurdles to Adoption

There are a number of gating factors to still consider. Deployment costs and data collection challenges remain significant hurdles for humanoid robots, compounded by limited dexterity and unstable connectivity. Effective operation requires robust real-time data processing and high-bandwidth, low-latency networks for over-the-air commands.

The widespread commercialization of humanoid robots faces numerous technological and societal obstacles. On the technical front, developing humanoids capable of navigating complex human environments will necessitate advancements in generative AI tailored specifically for these robots. Current cognitive capabilities are rudimentary, highlighting the need for enhanced "thinking" abilities.

Moreover, improvements in precision actuators, sensors, and battery life are essential to broadening the range of tasks humanoids can perform. Despite decades of robotics development, the rapid emergence of generative AI may lead to scenarios in which humanoids' cognitive capabilities are their physical capabilities, which could create potential hardware bottlenecks, malfunctions, or robots going completely rogue.

Social, policy, and safety considerations—similar to those of autonomous vehicles—pose additional challenges. The thing is, technology is advancing so fast that policy makers and regulators cannot even keep up. Any law that comes through is practically already “too late,” as the technology is already on to its next more advanced rendition.

The Moral Dilemma

Beyond technological hurdles, I think the other aspects of the risks are actually more philosophical and much harder to grapple with. When I saw the demo video for Galbot’s human maid, I was shocked to see that the child in the demo to throw the trash bin on the floor and running off while the robot replies with “I will pick up the trash”. As a mother of a toddler, I am putting much of my effort into teaching her mannerisms, how to behave, how to chew with her mouth closed, how to greet people with respect, and how to pick up after herself. Are we going to lose our humanity?

In the Unitree demo, the robot was also repeatedly kicked and pushed to test its durability and balance. Again, should we treat them that way?

My husband made fun of me when he saw how I was talking to Perplexity. I always start with “Could you please?” and often end my requests with “Thank you.” It’s not intentional, but it’s just how I’ve been wired to speak to people when asking them to do something. I unintentionally wrote like that to the machines. So, how should we be interacting with the robots? Can someone please provide a guideline? If not, at least some etiquette training or a set of norms to follow.

This also goes back to what I brought up in the Physical AI primer article: Where is the moral line between what cheating is and what is not? I was speaking to (who wrote a Physical AI piece here) a few days ago, and she said she saw a company called Exrobot (see instagram screenshot) where they were creating companion robots, aka robot girlfriends. How will people treat them? Will we see a robot edition of Planet of the Apes one day where the robots will revolt and keep track of a good list and naughty list of humans?

In the 2025 Predictions episode of the All-In podcast, Jason Calacanis said that there’s a psychological aspect of dominance and control when humans abuse the machines, it’s really to assert superiority and make sure they’re still in power. I mean, probably, that sounds logical but that’s pretty awful isn’t it.

I really don’t have the answers to these, but they’re making me increasingly wary of embodied AI and quite certain that I wouldn't want to bring one into my household; heck, I don't even like Siri or Alexa. Anyway, Diana and I will dive deeper into this topic another time to complement our initial feelings of shock and fear with more technical and scientific research.

Subscribe now

Conclusion, For Now

Looking at the two biggest players in robotics and AI, it’s the U.S. and China, once again. The United States excels in innovation and research, backed by a skilled workforce and proactive adoption of Industry 4.0 technologies. Yet, high integration costs and a lack of standardization have hindered broader implementation.

And conversely, countries like China can benefit from a robust manufacturing base and strong government policy-driven support, exemplified by initiatives such as "Made in China 2025," which aim to position the nation as a leader in robotics innovation. And we’ve seen the rise of Chinese robotics firms capturing a substantial share of the global market. However, challenges persist, including reliance on foreign technology and concerns about job displacement due to automation.

But no matter where you are, there is the issue that everyone faces, which is the undeniable job and thus societal disruption as automation becomes more prevalent, the ethical decisions we’ll need to make as humankind, and the safety risks we will inherently create and bear responsibility for our future generations.

Last hot take of the day: I think, in the space of Physical AI, first movers may not have the upper hand. Given how fast the technology is developing, there could be new players that completely leapfrog the current models available if they can integrate AI better into the mechanics of robotics. What we THINK is the next use-case might not be what sticks around.

This is part of a series on Physical AI. We’ve started with humanoid and quadruped robotics. And I have a few articles I’m working on with collaborators which will be on autonomous driving, EVs pivoting to AI-first companies, and then lastly, industrial robots (which are not as cool looking but definitely will be the most impactful on our labor force, if not already replacing many blue-collar workers). Actually, see ’s post here that talks about the consolidation in the warehouse automation space here.

“AI’s next stage is called physical AI. Physical AI is where AI interacts with the physical world. It means robotics.” – Jensen Huang, Nvidia, at a factory opening ceremony in central Taiwan on Jan 16, 2025.

Disclaimer: I’ve made some edits and fixed some typos, apologies. I’ve also added in new news on OpenAI’s robot plans to this article. This was originally published on Jan.17, and was updated on Jan. 20.

Subscribe now

This year, CES was all about AI integration into hardware, from wearables to AI-empowered furniture to robots! Stealing the thunder was not the non-sexy machinery robots that warehouses have all adopted—the fun ones we see in sci-fi movies—but the human-look-alike ones.

We can say robotics are having a ChatGPT ‘moment.” (as in the technology is not new, but the public and capital interest has experienced an awakening)

In particular, last year, we saw how Generative AI has drastically transformed robotics. Robots previously controlled by a remote can now “think” for themselves, integrating AI into the machine in a seamless hardware-software integration. We’re seeing the early days of two vastly different areas of science come together: LLM/Gen AI and robotics.

Just last week, it was reported that OpenAI secretly reopened its robotic division again after shutting it down in 2020. According to the company, their new robotics team will focus on “unlocking general-purpose robotics and pushing towards AGI-level intelligence in dynamic, real-world settings.”

[I will publish a deep dive into Unitree, the leading embodied-AI humanoid, and quadruped robotics firm, in a few weeks, so stay tuned.]

What I found helpful was to use this framework when trying to understand physical AI, to look at the whole thing as a continuous process of three steps:

• Sensing (the eyes, ears, and hands of the robot—sensory technology),

• Thinking (the decision-making process—GenAI),

• and Acting (the ability to follow through and have an impact in the physical world—mechatronics).

In a more technical way of categorizing this, , an AI writer, has said the system architecture can be broken down into:

1) Perception Layer: This is where the machine processes multi-modal sensory input through advanced sensor fusion algorithms. It helps the computer better understand the physical environment around it.

2) Cognitive Layer: This is the core of how AI integrates into traditional robots. It serves as the machine’s decision-making brain, processing the data collected from its perfection layer based on pre-trained simulations to decide how to react to the physical world.

3) Action Layer: This layer really creates the “output” to the physical world, ensuring accuracy in its movement and where we’ll see the jumping, walking, or object moving.

And physical AI encompasses so much: 1)autonomous vehicles (which I’m working on a piece), 2)Specialized Robots - warehouse/ labor automation (traditionally what we know of), and 3) Humanoid Robots.

Disclaimer: Today, we will only focus on Humanoid robots. I tried my best to make my thoughts logical as I jotted them down and attempted to understand the evolving relationship between robots and AI.

Share

Humanoid Robots

First, what is embodied AI?

Embodied AI refers to the artificial intelligence (AI) agents — robots, virtual assistants, or other intelligent systems — that can interact with and learn from a physical environment. (Qualcomm’s definition)

For many years, we have gotten used to robot arms for warehouse use, coffee-making robots all over Tokyo and Shanghai, robot dogs for fun and delivery purposes, and other such devices. But these robots kind of lost their wow factor.

About ten years ago, a family friend invested in a robot restaurant in Beijing. It was a machine that could chop and make like 10 dishes that were programmed into the machine. My mom thought it was so cool, but the robot didn't look like a human, just another machine, so the novelty didn't work on me (I'm pretty sure they closed the shop after a year). The dishes tasted fine, but cooking isn't science; you need to feel the fire and how much salt to add. Sometimes, you need to jazz it up and improvise. The more suitable use case is for a big university/ campus cafeteria, factories, or prisons where you need to provide standardized food on a mass scale efficiently. It just didn't work as a mid-range priced restaurant in the CBD of Beijing. There are many other better options to eat there than a meh-tasting robot-made food.

That, was not embodied-AI. That was just a pre-programmed robot.

Now, let’s dive deeper into humanoids, robots that are physically designed like humans. They have two arms, two legs, and a head, and their mobility mimics that of our bodies.

And I agree with him. Our obsession with human-looking robots is now beyond how egotistic and self-important we are as humans. It’s an optics thing but also a practical thing. So now it’s sounding more and more like I, Robot.

The mainstream interest in humanoid robots started in late 2022 when Telsa unveiled its Optimus. Since then, many robotics companies have received attention from the media and investors worldwide. This became the new “wow” robot form.

How are they different from traditional robots?

Traditional robotics has been around for over a decade. These conventional robots relied on rigid programming to execute pre-written instructions (following a set of orders and could not “think”). Traditional robots were essentially limited to repetitive, pre-programmed motions with little flexibility, and the complexity of tasks they can handle is constrained. Whereas embodied AI robots are trained on real-world data that uses reinforcement learning and they have developed the ability to “think.”

In October 2024, during a video call at a tech conference in Riyadh​​, Elon Musk predicted that by 2040, there would be at least 10 billion humanoid robots priced between $20,000 and $25,000 through a video call at a tech conference in Riyadh. While we are still far from that, we are one step closer to robots becoming more accessible/ prevalent with the frontier robot models from Ameca, Tesla, Agility Robotics, Unitree, and so on.

In his most recent AGM, Musk, the Tesla CEO, predicted that humanoids will outnumber humans by two to one or more in the future. That is wild (and scary).

And just today, David Soloman, CEO of Goldman Sachs, said that AI will revolutionize business productivity. His view is that eventually, much of what the junior bankers do will be replaced by AI, such as drafting 95% IPO filing documents, which can now be completed by AI in minutes, “the last 5 per cent now matters because the rest is now a commodity.”

Executives are looking at this revolution and really thinking that AI robots will replace (assist) us one day.

Development and Limitations

Nvidia’s keynote speech in March 2024 also focused on physical AI- robotics. The advancement of robots relies on three main factors: 1)GenAI, 2) actuators and mechanics, and 3) battery storage (which I’ve also touched on before when writing about renewable energy solutions here)

If the three factors can progress together, the mass scaling and commercialization of humanoid robotics will likely be in a few years. Now, in the Unitree piece, I talk about China, given its long experience in hardware manufacturing, which has given it a know-how advantage in mechatronics, which combines mechanical technology and electronics, given it being the hardware manufacturing hub for various appliances over the last few decades. While the U.S. is still obviously leading in GenAI, the question is which factor matters more, or will there be an intention for collaboration? And who can figure out the last piece? A better battery solution. Can they make batteries lighter, last longer, and be safer?

We humans have been dreaming of a world where robots are side-by-side with us for decades, from the movie Her in 2013, where we imagined a virtual AI lover (where that can practically be real now), to the 2024 sci-fi thriller featuring Megan Fox Subservience where physical AI challenges our morals of faithfulness, loyalty and even leads to detrimental safety risks. The imagination has always been that we will have robotic selves created one day. It’s just that now, the technology has finally advanced enough to catch up to our imagination. (obviously, the current humanoids look nothing like Megan Fox yet)

So, to my point, this is obviously just the beginning. As resources now all go to physical AI (or AI in general), this interdisciplinary field will mature out of its nascent stage and continue to advance faster than maybe we can even comprehend. As our physical and digital worlds continue to mold and blend into each other, the safety concerns really worry me as a mother of a 2-year-old. If smartphones are already confusing children about reality, how should we regulate this technology, and what will robots be capable of when she reaches her teenage years?

There is still a lot to unpack, so I will add a new vertical to AI Proem: Physical AI. In this vertical, I will explore the many areas that may affect the development of Physical AI. First, I will do a deep dive into China’s leading humanoid robot company, Unitree Robotics, which is often compared to Boston Dynamics. I will also explore the battery solution angle, which I have been trying to find time to look into after my renewable energy piece.

Then, I think it will be interesting to explore the concerns around possible theories on how robots will go rogue. This concerns overall robotics safety and the dexterity of robot machines. Then, there are the technicalities of connectivity, bandwidth, latency complications, and other key components—sensors (LIDAR), sensor fusion, and software architecture technology. This is also related to another work-in-progress piece that will examine how EV companies are transitioning to AI-first companies following Tesla’s footsteps (Li Auto, for example).

I also want to expand my coverage to include how AI will affect our workstream, so I am exploring how our labor force will be impacted by physical AI adoption and potentially then investigating and coming AI software or how AI will affect our most popular internet-era software platforms.

There is just so much to learn and write about! For now, toodeloo~