On Feb. 16, hundreds of millions of households watched as humanoid robots from four different Chinese companies danced, acted in a comedy skit, did parkour, and performed martial arts onstage at the Spring Festival Gala, China’s most-watched television broadcast. Across the country, drone shows lit up the night skies as China celebrated Lunar New Year, the synchronization of tens of thousands of drones coordinated by artificial intelligence.
[time-brightcove not-tgx=”true”]The physical AI fervor has traveled across the Pacific. At the glitzy Consumer Electronics Show in Las Vegas earlier this year, Chinese startups dominated the convention with AI-enabled hardware from smart home appliances and wearables to all kinds of robots.
While American frontier labs are battling each other across large language model leaderboards, China’s AI capabilities are showing up in physical ways—leaving screens and entering our daily lives. We’ve lived through over a decade of, in the words of venture capitalist Marc Andreessen, “software eating the world.” Now, metal and mathematics have converged and hardware is eating the world.
As AI becomes integrated into our physical world, we’re hurtling into a new chapter of embodied intelligence. Unlike the past few years, where China has been playing catch-up in AI models, China is pulling ahead of the U.S. in physical AI.
China’s all-in on physical AI
Imagine streets filled with robotaxis and delivery robots. Imagine general-purpose humanoids becoming as ubiquitous as smartphones, doing household chores, taking care of our parents, and taking over tasks that humans don’t want. Imagine 24/7, fully automated “dark factories” producing more robots without human workers. And imagine nations commanding drone swarms and robot dog packs capable of making decisions and executing missions without human intervention.
We are on the cusp of this future—and much of it is currently driven by China. For a long time, a main bottleneck for robotics has been scalability. Now, hardware costs have more than halved as China achieved manufacturing dominance in adjacent industries like electric vehicles, which have spurred innovations and economies of scale in components that overlap with robots, including actuators, sensors, and batteries. Meanwhile, recent advancements in multimodal AI (models that can process multiple types of information, like visual and audio at once) are improving robots’ generalization capabilities for everyday tasks.
Today, China controls much of the supply chain that underpins robotics. It’s the world leader in lidar sensors, with an estimated 70% of the global market. Suzhou-based Leaderdrive has quickly become one of the world’s biggest producers of harmonic reducers, gears which are crucial to a wide range of robotics. Eyou Robot Technology recently opened the world’s first automated production line for humanoid robot joints in Shanghai. And Chinese companies like ESTUN and Inovance are emerging as dominant players in controllers, which essentially serve as a robot’s brain.
Humanoid robots
China’s ability to domestically manufacture at scale has driven down the costs of robots, including humanoid forms, which can be extremely expensive to produce due to the sheer versatility and dexterity of human movement.
Last year, Chinese firms released several entry-level humanoids for home use to the mass market, including Noetix’s Bumi, a family companion and education robot that costs as little as $1400.
While the technology needed for adaptive humanoid robots is not yet fully mature, whichever country deploys robots faster will collect more data, which in turn unlocks better deployment. In 2025, China accounted for over 80% of global humanoid robot installations and over half of the world’s industrial robots. What’s more, cities across China such as Beijing, Wuhan, and Shanghai are opening training sites for acclimating robots to various settings (including environments that mimic a retail outlet, an elderly care facility, and a smart home) and harvesting standardized data.
The fervor for still-nascent humanoids is a sign of what’s to come for physical AI. Since as early as the 1970s, companies have been trying to build autonomous robots, but were stuck at pre-programming robots for fixed tasks that only worked in controlled environments. After decades of attempts, however, the generative AI revolution is now making it possible for robots to perceive and sense the real world, generalize from limited training data to novel situations, and thus learn to operate in dynamic environments.
In time, robots will be able to reason, adapt, and execute in real time without needing a constant cloud connection. Eventually, robots will be able to work, build, manufacture, farm, and fight on battlefields without humans. Robots will be a labor force that needs no rest. Robots will automate entire supply chains and perform physical tasks that humans cannot do. And China is currently winning the robotics revolution.
The future of the AI race
To be sure, there is still reason to believe that the U.S. can regain the lead in physical AI.
We would argue that China likely suffers from excessive competition and waste (in 2025, the country tallied over 150 humanoid robot startups). And while China’s humanoid robots are glitzy, many can’t reliably do skilled human tasks. Plus, there’s little consensus on when we’ll see them enter our daily lives. The world’s second-largest economy is also still dependent on foreign suppliers for certain high-end components like advanced servo motors. Meanwhile, the U.S. leads in advanced simulation platforms and has notable players like Tesla, Figure AI, and Physical Intelligence, which may still outpace Chinese competitors by focusing on software breakthroughs over sheer hardware volume.
Yet, China’s bot playbook is reminiscent of how it built its lead in the EV industry: early state support enabled a myriad of entrants, created more demand, boosted production volume, and cultivated manufacturing experience, which resulted in intense competition and economies of scale that fostered globally competitive brands.
When it comes to physical AI, the U.S. has distinct advantages in software, foundational research, talent, and chips that will prove indispensable. But in the age of hardware eating the world, America also needs to learn from what China has done. The U.S. needs to nurture the sector, rebuild its supply chain with allies that produce critical components, support open-source models to accelerate robot development, possibly regain manufacturing expertise through reverse tech transfer and joint ventures with Chinese firms, and deploy American robots in certain sectors that can act as a sandbox—starting first on factory floors.
The Chinese robots are coming, whether or not America builds its own.
