The U.S.- China Robot Race 🤖

On April 19, 2026, amidst clear skies and a mild breeze, nearly 13,000 runners participated in a half-marathon through China’s Beijing Economic-Technological Development Area, famously known as the E-Town.

And the winner of the half-marathon was a participant named Lightning. Staying true to the name, Lightning completed the 21-kilometer race in 50 minutes and 26 seconds — surpassing the current world record of 57 minutes and 20 seconds, previously held by Uganda's Jacob Kiplimo, which he set in Lisbon just a month before this race. 

Now, that sounds like a great athletic feat. And in a sense, it is. Except it isn't just an athletic feat. It's a technological one as well. 

Because Lightning is not human, it is a humanoid robot developed by the Chinese company Honor!

While the world has spent the last few years watching the US-China rivalry play out in semiconductors, in AI models, and in data centers, and with everyone fixated on those headlines, a quieter race has been taking shape — one fought not in server farms or fab labs but on machines that behave like Homo sapiens. 

Today, humanoid robotics is emerging as the next great battleground between Washington and Beijing. And if the E-Town Marathon was any indication, China isn't just competing. It's already running fast and making remarkable strides. 

To understand the race between the US and China to build humanoid robots and their future, let's start with what humanoid robots are and their history. For that, we need to go back to 400 BCE. 

A humanoid robot is an autonomous machine built in the shape of a human head, torso, limbs — designed to move through the same spaces we do, use the same tools we use, and do the same work we do.

In simple terms, it is a machine built to look and act like a human. 

Interestingly, the story of humanoid robots doesn't begin in a lab; it begins in the human imagination, in myth and religion, thousands of years before anyone had the tools to build one.

The ancient Greeks were dreaming about it as far back as 700 BCE. In Homer's Iliad, the god Hephaestus — Greek god of fire and metallurgy — forged golden handmaidens, with human voices, to serve as his assistants. 

In the east, China had its own version. In an ancient Chinese text from the 4th century BCE, a craftsman named Yan Shi presented King Mu of Zhou with a life-sized human figure built from leather and wood. It could walk, sing, and move every part of its body.

These were all fragments of imagination. It would take nearly 2,000 years for these ideas to take any real shape.

The first person to actually build something close was the Arab polymath and engineer, Al-Jazari. In 1206, he published The Book of Knowledge of Ingenious Mechanical Devices — a manual that described, in precise detail, programmable automata. One of his creations was a boat carrying four mechanical musicians that could play music automatically, powered by water. 

Two centuries later, Italian Renaissance artist Leonardo da Vinci went further. Around 1495, he sketched detailed plans for a mechanical knight — a suit of armor driven by a system of cables and pulleys, capable of sitting, standing, and moving its arms and jaw.  

In fact, during the Renaissance, multiple artists and thinkers dabbled with the idea of automata — self-operating machines designed to mimic the movements of living beings. 

But interestingly, all these years, none of these machines were called robots. That word didn't exist yet. It was only in 1920 that Czech playwright Karel Čapek introduced it to the world, derived from the Czech word robota, meaning forced labor.

In the early 20th century, engineers across Britain and America were building mechanical humanoid robots for spectacle. Britain's Eric could stand, move, and deliver speeches. America's Elektro could walk, talk, smoke cigarettes, and draw a crowd of thousands. 

They were marvels of engineering. But they lacked the one thing that would make a robot truly useful: intelligence. The engineers who followed — in Britain, in America, in Europe — couldn't crack it either. The breakthrough, when it finally came, arrived from an unlikely direction: a university campus in Tokyo.

The 1970s were a transformative decade for Japan. In just two decades after the war, Japan had become the world's second-largest economy. By the 1970s, it wanted to grow out of its industrial economy into a more stable, consumer-oriented one — with its eyes firmly on the future.

And it was building that future at Waseda University.

It was here in 1973, a team of engineers and researchers led by Professor Ichiro Kato unveiled WABOT-1 — the world's first full-scale humanoid robot. It could walk, albeit slowly. It could grip objects with its hands. It could communicate in Japanese. 

At that point, it was the most advanced humanoid robot on the planet. But it was also deeply limited — it moved at a crawl, and its cognitive capacity was roughly that of an eighteen-month-old child. Two problems, in other words: the body and the mind.

Honda spent the next three decades solving the first one.

In 1986, Honda launched a classified internal research program. Its engineers spent years studying human movement obsessively. They weren't building a machine. They were reverse-engineering a human being.

And in 2000, the result of that work walked onto a stage. Honda unveiled ASIMO — the Advanced Step in Innovative Mobility. It could climb stairs, recognize faces, respond to voice commands, and run at 9 kilometers per hour. It soon became the face of humanoid robots worldwide. 

But ASIMO never became a product. Neither did any of the humanoid robots that followed it out of Japan's labs. Similarly, in the United States, Boston Dynamics' Atlas received the same reception. It was unveiled in 2013 to amazement and hope, watched by millions, but never graduated beyond the experimental phase into anything resembling a commercial product.

The problem, for everyone, wasn't capability. It was the cost. A single ASIMO unit costs millions of dollars to build and maintain. Atlas wasn't much different. The sensors, actuators, and computing power required to keep a two-legged machine upright and functional were beyond what any commercial market could absorb. 

But there was another reason humanoid robots never crossed the commercial threshold. They had no advanced intelligence. 

ASIMO had cracked the body. But the mind — the ability to adapt, learn, and make decisions in real time — remained unsolved. 

However, in the early 2020s, two things happened at once, and the humanoid robotics landscape changed entirely

In November 2022, OpenAI released ChatGPT. Within days, the world understood something fundamental had changed. 

Large language models — trained on vast amounts of human text — could reason, adapt, and respond in ways no machine had before. For robotics, the implications were immediate. For decades, the limiting factor for humanoid robots had been the mind: the ability to process a real-world environment, make decisions, and adapt in real time. Large language models, combined with rapid advances in computer vision, suddenly offered a path to solving exactly that.

The second was affordable tech. But the seeds of that were sown a decade before, in China.

In the early 2000s, after China joined the World Trade Organization in December 2001, it soon became the manufacturing capital of the world. From just 4% of world exports in 2001, China's share climbed to nearly 16% by 2024. 

But the manufacturing was low-cost, low-margin, and deeply exposed. China was the world's factory floor — not its engineering lab.

And Beijing understood, clearly, that assembling the world's smartphones and stitching its garments was not a foundation for lasting power. 

So it decided to climb the value chain. In 2015, the government launched Made in China 2025 — a sweeping industrial strategy targeting ten high-technology sectors, from semiconductors to aerospace to new energy vehicles. Its goal was unambiguous: reduce reliance on foreign technology and move China up the global value chain.

Chinese firms — from CATL in batteries to BYD in drivetrains — didn't just learn to make these components. They learned to make them at scale, at speed, and at costs that no other country could match. 

Then the robotics wave arrived.

When Chinese firms began pivoting to humanoid robots in the early 2020s, they looked at their EV supply chain and realized something remarkable: they already had almost everything they needed. The motors that drove EV wheels could drive robot joints. The battery cells that powered cars could power bipedal machines. The sensors built for autonomous driving could teach a robot to perceive its environment.

No other country had built this combination by accident. And no other country could replicate it quickly. China wasn't just positioned to compete in humanoid robotics. It was positioned to own it – and it did exactly that. 

As of early 2026, China has emerged as the global leader in humanoid robot production, with over 150-200 companies focusing on embodied AI and robotics.

According to Omdia, global humanoid robot shipments reached approximately 13,000 units in 2025 — nearly five times the volume shipped the year before. Chinese manufacturers accounted for nearly 90% of global humanoid robot shipments in 2025, with the bulk of that output coming from four companies: AgiBot, Unitree Robotics, UBTECH Robotics, and Xiaomi. AgiBot shipped more than 5,100 humanoid robots during the year, while Unitree Robotics claimed it had shipped over 5,500 units of its own.

The contrast with American competitors was stark. Tesla shipped only a few hundred Optimus units during the same period. Figure AI and Agility Robotics each delivered roughly 150 robots. 

In fact, it is not just the robots themself, but if you look inside the robots, the Chinese domination is even clearer. 

According to McKinsey's April 2026 report on the humanoid supply chain, a robot's bill of materials is broken down into five hardware domains. 

Actuators — the motors that drive every joint in the body — account for 40 to 60 percent of total unit cost. Sensing and perception systems take 10 to 20 percent. Compute and control platforms 10 to 15 percent. Structural components and battery modules 5 to 10 percent each. Together, those five domains account for 85 to 90 percent of the cost of building a humanoid robot.

Here’s where it gets interesting: China dominates the manufacturing of almost all these parts. 

In addition to controlling approximately 90 percent of permanent magnet processing globally — the rare-earth magnets that power every actuator in every humanoid — China holds 40 percent of precision bearing capacity, 35 percent of motors, and 30 percent of power electronics. These are not peripheral inputs. They are the physical core of the machine. 

Which means that even if another country decides to invest heavily and build humanoid robots at scale, much of the underlying supply chain would still run through China. So while companies in the United States or Europe may design the robots, a significant share of the components powering their joints, movement, sensing, and control systems would still trace back to Chinese manufacturing capacity.

Speaking in a podcast interview in March, Jensen Huang, CEO of NVIDIA, acknowledged just how deeply China dominates the industrial foundations of robotics.

And the Chinese government does not want to let go of this dominance.

In 2023, China made that ambition explicit when Beijing announced plans to build a resilient domestic humanoid robotics supply chain by 2027 — specifically to reduce dependence on foreign suppliers such as Japan and Germany for critical robot components.

It has now taken it one step further.

Humanoid robotics and embodied AI are now being folded directly into China’s 15th Five-Year Plan (2026–2030), where Beijing has identified them as strategic “industries of the future” alongside quantum computing, 6G, and brain-computer interfaces.

Under the plan, China aims to accelerate the deployment of AI-powered robots across manufacturing, logistics, healthcare, and other industrial sectors as part of a sweeping national “AI+” strategy.

That strategy is not just about automation. It is about reshaping China’s industrial competitiveness for the next phase of global manufacturing.

In a recent one-on-one interview with CrossDock Insights, Dr. Amitendu Palit explained that embodied AI is expected to serve two strategic functions for China. “Embodied AI is going to play two roles. One is increasing the productivity of traditional industries. For example, in textiles, and partly in automobiles and chemicals, embodied AI will increase competitiveness across various levels of the supply chain.

Second, AI will be used to make China the first-generation leader in strategic industries of the future. Biomedicine would be one such. Advanced manufacturing would be another,” he said. 

While China has been making significant strides in humanoid robot manufacturing, what exactly has the United States been doing in response? Let’s break it down. 

While China has been scaling humanoid robot manufacturing through state-backed industrial policy, the United States has largely relied on private companies and venture-backed startups to drive the sector forward.

At the center of that push sits Tesla. 

In August 2021, Elon Musk walked onto a stage at Tesla's first AI Day and announced that the company was building a humanoid robot. To illustrate how early the idea was, a person in a robot suit came out and danced. Nobody took it seriously. Tesla was a car company.

A year later, an actual robot walked onto that same stage. 

Elon Musk has repeatedly framed humanoid robots as central to Tesla’s long-term future. In a 2024 post on X, Elon Musk wrote that “~80% of Tesla’s value will be Optimus,” signaling just how strategically important the humanoid robot project had become for the company.

Around the same period, Musk also predicted that Optimus could eventually turn Tesla into a $25 trillion company — a valuation that, at the time of his comments, was worth more than half the total market capitalization of the entire S&P 500

What followed was four years of grinding development and slipping deadlines. 

Tesla had originally promised a commercial Optimus working inside its own factories by the end of 2025. That didn't happen. In January 2026, on the Q4 earnings call, Musk admitted that no Optimus robots were yet doing useful work at Tesla. The company is now on its third generation of the robot, with mass production at its Fremont factory scheduled for late July or August 2026.

According to Elon Musk, consumer availability could arrive by the end of 2027 — provided the robot clears the reliability and safety bar Tesla has set. The long-term target price is $20,000 to $30,000 per unit.

Across Silicon Valley, a growing wave of AI and robotics startups is racing to build humanoid robots — and many are backed by some of the biggest names in Big Tech.

At the center of that ecosystem sits Figure AI, one of the most heavily funded humanoid robotics startups in the world.

In 2024, Figure raised $675 million from a consortium that included Microsoft, NVIDIA, OpenAI, Amazon founder Jeff Bezos, and other major investors, pushing the company’s valuation to $2.6 billion. 

In October 2025, Figure unveiled Figure 03 and announced plans to ship 100,000 units over the next four years through BotQ, its dedicated humanoid manufacturing facility in California. 

Then there is Apptronik, the Texas-based humanoid robotics startup. Unlike many robotics startups still operating largely in demo mode, Apptronik is already testing humanoid robots inside real industrial environments.

Early Apollo units are currently operating within designated sections of factories and warehouses run by partners including Mercedes-Benz, GXO Logistics, and Jabil, where the robots are being trained to transport components, sort materials, and support repetitive industrial workflows.

And investors are pouring money into the company.

In early 2026, Apptronik raised $520 million at a $5 billion valuation in a funding round co-led by Google-backed investors, bringing its total Series A funding to roughly $935 million.

Next in the list attracting major attention is 1X Technologies, the humanoid robotics startup backed by OpenAI and NVIDIA. In 2026, 1X launched what it described as America’s first integrated humanoid robot factory in California, with plans to manufacture more than 100,000 NEO humanoid robots annually by 2027

Another major trend is that Big Tech companies no longer want to remain on the sidelines of the humanoid robotics race. Meta is one of the clearest examples.

In May 2026, Meta acquired humanoid robotics startup Assured Robot Intelligence (ARI), a company focused on building foundation models that allow robots to understand, predict, and adapt to human behavior in real-world environments.

However, there is a catch.

Even when many of these humanoid robots are being designed by American companies, a significant portion of the components inside them — from motors and actuators to rare-earth magnets, batteries, sensors, and power electronics — still trace back to Chinese supply chains.

In other words, the software may be American.

But much of the industrial hardware underneath the robot remains heavily dependent on China. And that dependency is increasingly making policymakers in Washington uncomfortable.

U.S. officials have grown concerned that Chinese firms are already positioning themselves at the center of the global humanoid robotics supply chain, especially in components that could eventually have military, surveillance, or other sensitive dual-use applications.

In February, a bipartisan group of lawmakers in the U.S. House of Representatives proposed legislation to establish a national commission focused on America’s competitiveness in robotics, specifically citing manufacturing weaknesses, supply chain vulnerabilities, and rising dependence on foreign suppliers for critical robotics components.

So why can’t the United States simply manufacture all of these components itself?

Part of the answer is technical expertise and industrial capacity. China has spent decades building dominance across precision manufacturing, rare-earth processing, motors, batteries, magnets, and industrial electronics. All of which have helped them make better robots.

And another major factor is cost.

Humanoid robots are extraordinarily supply chain-intensive machines, and China’s manufacturing ecosystem currently operates at a scale and efficiency that few countries can match.

According to research from Morgan Stanley, a humanoid robot built using Chinese components costs roughly $46,000 to manufacture.

The same robot, built using non-Chinese supply chains, costs approximately $131,000 — nearly three times as much. For Tesla, which aims to eventually bring the cost of Optimus down to roughly $20,000 per unit, reducing dependence on China’s supply chain may prove extraordinarily difficult.

That cost gap is enormous and simply not feasible. But why should the U.S. be worried about an industry that seems to be a relatively small industry, when compared to AI or semiconductors? 

Today, the humanoid robotics industry remains relatively small.

Barclays estimates the global market at just $2 billion to $3 billion today. Barclays believes the sector could grow to roughly $40 billion by 2035 — and potentially as high as $200 billion under more optimistic scenarios.

Morgan Stanley projects an even larger number, estimating that the broader humanoid robotics ecosystem — including manufacturing, supply chains, maintenance, and services — could eventually exceed $5 trillion by 2050.

And beyond economics, there are also strategic concerns.

Humanoid robots could eventually play important roles across manufacturing, logistics, elder care, warehouses, defense support systems, and other critical industries. That possibility is one reason governments and major technology companies are already treating the sector as strategically important despite its early stage.

But not everyone is convinced the hype will fully translate into industrial reality.

In a recent conversation with CrossDock Insights, Harvard Business School Professor Willy Shih questioned whether humanoid robots make economic sense for many industrial applications.

“From an industrial automation standpoint, you add a lot of cost by having all those mechanisms and control systems associated with maintaining balance, associated with the so-called humanoid functions,” Shih said.

Shih argued that many industrial environments may not actually require robots that resemble humans. “We saw the humanoid robot showcase at the Beijing Olympics, we saw them playing table tennis, and that is all great,” he said. “But from an industrial standpoint, do I need that additional function that would warrant the additional cost?”

Humanoid robotics is rapidly becoming the next great strategic frontier — as important as semiconductors were in the 1980s, or as artificial intelligence is today. The United States built the internet and largely shaped the first era of AI, but it missed much of the manufacturing revolution. It cannot afford to miss this one.

The uncomfortable truth, however, is that China is miles ahead in this race. But the United States still has the software, the capital, and the ambition to catch up in the race.

And one thing is becoming increasingly clear: humanoid robots running marathons, moving through warehouses, picking pallets, operating inside factories, finishing household chores, and even supporting military operations are no longer ideas confined to science fiction. Slowly but steadily, they are beginning to enter the real world. 

This newsletter was written by Shyam Gowtham