China’s humanoid robots and the new speed‑to‑scale arms race
China’s humanoid robots are no longer just viral clips doing kung fu flips on TV; they’re quietly turning into a test case for how fast a country can industrialize a new general‑purpose technology. While Western audiences debate whether humanoids are hype, Chinese firms are already shipping tens of thousands of units, backed by a state that treats robotics as infrastructure. In this piece we’ll look beyond the showreels: why China is winning the early humanoid market, what bottlenecks still limit its dominance, and what this means for European and global players who risk waking up in a world where the robot supply chain is effectively made in China.
The news in brief
According to TechCrunch’s reporting, China’s humanoid robot industry has moved from attention‑grabbing demos to meaningful early deployments. Robots from companies like Unitree, Galbot, Noetix and MagicLab shared the stage at China’s massive Spring Festival Gala TV show, while smartphone maker Honor is preparing to unveil its first humanoid at MWC in Spain.
Humanoids sit squarely inside Beijing’s industrial strategy. Robotics was a pillar of the “Made in China 2025” plan, originally focused on factory automation. Now, rapid progress in multimodal and “embodied” AI is pushing China to treat humanoids as a way to counter labour shortages and boost productivity.
TechCrunch cites analyst Selina Xu, who notes that China’s hardware supply chain—largely built around EVs, batteries and sensors—lets domestic humanoid makers iterate faster and cheaper than U.S. rivals. Chinese firm Unitree reportedly shipped around 36 times more units last year than U.S. players Figure and Tesla.
A Forbes report referenced by TechCrunch estimates global humanoid shipments at just 13,317 units last year, with a projection of 2.6 million units by 2035. Most of today’s volume comes from Chinese players such as Agibot, Unitree, UBTech, Leju Robotics, Engine AI and Fourier Intelligence.
Why this matters
The key fact is not that China has humanoid robots; it’s that it can scale them faster than almost anyone else. This is the same formula we saw in solar, batteries and EVs: once the product category is “real but immature,” China’s combination of policy support, capital and manufacturing density turns into a speed‑to‑scale weapon.
The immediate winners are Chinese robotics startups and their domestic supply chain: motor makers, sensor vendors, battery producers and contract manufacturers that can seamlessly repurpose EV know‑how for bipedal machines. Their robots don’t have to be best in class to win the early market; they just have to be good enough and available in volume.
The losers—at least in the short term—are Western firms that treated humanoids primarily as flashy R&D platforms. When a Chinese vendor can offer a robot that’s 30–50% cheaper, delivered this quarter rather than next year, operations managers in logistics, retail or light manufacturing will at least pilot the Chinese option.
TechCrunch’s sources describe a shift from “demo‑driven excitement” to “operations‑driven adoption”. That’s crucial. Once robots prove they can reliably unload boxes, patrol warehouses or run night‑shift tasks, procurement logic takes over. Price per hour and mean time between failure matter more than who has the slicker YouTube video.
However, China’s lead is narrower in software and autonomy. The article underlines that current humanoids are strongly limited by data and safety constraints; most companies still depend on Nvidia’s stack and heavy simulation. This is the window of opportunity for non‑Chinese players: if you can own the “robot brain” and the operating system layer, you can still shape the market even if much of the hardware is assembled in Shenzhen.
The bigger picture
Humanoids sit at the intersection of three megatrends: industrial robotics, generative AI and demographic decline. China happens to be under pressure on all three.
First, its workforce is ageing and shrinking. For a country that built its prosperity on labour‑intensive manufacturing, humanoids are a way to protect export capacity without opening the door to mass immigration. Japan is following a similar logic, which is why TechCrunch highlights its longstanding robotics tradition and focus on eldercare robots.
Second, generative and multimodal AI are finally good enough to control complex bodies, at least in constrained environments. We are essentially watching the birth of a “GPT for the physical world” — vision‑language‑action models and world models that can reason about what the next physical state should be. The TechCrunch piece correctly notes that this is still early: unlike language models, you can’t scrape the whole internet for robot motion data, so companies lean heavily on simulators.
Third, the hardware story rhymes with past industries. Think back to solar panels or lithium‑ion batteries: Western labs pioneered many core technologies, but China industrialized them. Humanoids are replaying this pattern. U.S. firms like Tesla (Optimus), Figure and Hyundai’s Boston Dynamics focus on high‑performance robots and tight factory integration; Japanese companies emphasise precision and control. China optimizes the full stack from gigafactory‑grade actuators to mid‑range humanoids you can deploy in a mall.
If Nvidia is currently the “Intel + Windows” of humanoids—providing the dominant chips and software stack—the open question is whether any country or company can build a genuine alternative before export controls and supply tensions bite harder. Chinese startups may be physically assembling many of the early robots, but their dependence on U.S. silicon and IP is still a strategic vulnerability.
The European / regional angle
For Europe, this story is uncomfortably familiar. We have world‑class industrial automation (ABB, Siemens, KUKA), strong research labs and a deeply robotics‑friendly manufacturing base—especially in Germany, Italy and Central Europe. We even have niche humanoid players, such as PAL Robotics in Spain. Yet we are largely spectators in this first humanoid scaling wave.
Two forces pull in opposite directions. On one side, the EU AI Act, product liability rules and the upcoming Machinery Regulation will make it harder to deploy unsafe humanoids in European workplaces. From a safety and workers’ rights perspective, that’s good. From a speed‑to‑market standpoint, it’s a handicap compared with China, where pilot deployments can be rolled out with less friction.
On the other side, Europe has a structural need for automation. Ageing populations in Germany, Italy, Spain and much of CEE, combined with chronic labour shortages in logistics, care and hospitality, mirror the pressures in China and Japan. European employers will want humanoids—if not in 2026, then certainly over the next decade.
The risk is repeating the solar panel story: European firms pioneer standards, ethics and early products, while Chinese and possibly Japanese and Korean vendors dominate mass production. If the default warehouse humanoid in Poland or the Netherlands ends up being Chinese‑made, Europe will import not just hardware but also embedded software, data flows and potential dependencies.
The opportunity is to treat robotics more like telecoms: invest in local champions, shape global safety and interoperability standards, and insist that robots deployed on European soil comply with strict data governance and cybersecurity rules. Europe may never match China’s unit volumes, but it can absolutely set the rules for what a trustworthy humanoid looks like.
Looking ahead
Over the next three to five years, expect humanoids to remain mostly “contained‑environment” workers. TechCrunch’s sources anticipate early traction in industrial manufacturing, warehouse logistics and retail: repetitive, long‑hour tasks with clear processes. These are ideal training grounds for robot brains and for debugging safety.
China’s advantage will likely compound: domestic policy, government procurement and private capital are all aligning to subsidise large‑scale pilots. Every deployment generates data, and every data point makes the next generation of models slightly better. Western firms, by contrast, will have to fight harder for access to real‑world environments at scale.
At the same time, several things could slow China’s march. A high‑profile accident could trigger public or political backlash, especially if humanoids are rolled out too aggressively in public spaces. U.S. export restrictions on advanced chips or AI tooling could bite harder, forcing Chinese firms to rely on less capable domestic silicon. And global buyers—particularly in Europe—may be wary of over‑reliance on Chinese robotic platforms for critical infrastructure.
Watch for three signals:
- Who defines the de facto “robot OS”? If Nvidia’s stack becomes as dominant in humanoids as Windows once was on PCs, software‑centric players may hold more power than hardware assemblers.
- Where do the first truly autonomous deployments happen? The country that proves safe autonomy in messy, semi‑public environments will gain enormous soft power and data advantages.
- How do regulators react to the first incidents? Europe’s AI Act, China’s emerging AI rules and U.S. liability debates will shape not just safety baselines but also the cost of doing business.
For European companies, the tactical move now is to start experimenting—with sandboxes, pilot projects and local partnerships—rather than waiting for a perfectly regulated, risk‑free humanoid. Those who build operational know‑how early will be far better positioned when the technology matures.
The bottom line
China is winning the early humanoid race for the same reason it won in batteries and solar: speed‑to‑scale backed by industrial strategy. But the game is far from over. Autonomy, safety, software ecosystems and regulation will decide who controls the “brains” of tomorrow’s robots. Europe can either become a passive importer of embodied AI or a standards‑setter and demanding customer that shapes how humanoids integrate into work and society. The question is whether policymakers and industry leaders move fast enough—this time—to matter.
