China Just Did to Humanoid Robots What It Did to EVs — and America Isn't Ready

The moment the EV playbook clicked into place, you could almost hear it. In 2015, Western automakers dismissed Chinese electric vehicles as cheap copies destined for domestic markets. By 2023, BYD had surpassed Tesla in quarterly sales. By 2025, Chinese EV brands controlled nearly 70% of their home market and were expanding aggressively into Southeast Asia and Europe. Now, almost frame-by-frame, the same sequence is beginning in humanoid robotics , and the data is in.

In 2025, Chinese humanoid robot startup AgiBot shipped an estimated 5,168 units, capturing 39% of the global market and claiming the top spot in Omdia's first comprehensive humanoid robot shipment rankings. The next five positions in that list? All Chinese. Figure AI and Tesla , the two US companies that made the top 10 , each sold approximately 150 robots. The gap is not a rounding error. It is a structural divergence that will compound with every passing quarter, and the US robotics industry is not yet having the honest conversation about what it means.

What Actually Happened

The global humanoid robot market shipped approximately 13,000 units in 2025, according to Omdia , the research consultancy whose rankings have quickly become the industry's primary measurement benchmark. AgiBot, officially named Shanghai AgiBot Innovation Technology Co., led all producers with 5,168 units. Unitree Robotics placed second, followed by UBTech Robotics Corp. in third. All six top positions were held by Chinese firms. The two US companies that cracked the top 10, Figure AI and Tesla, did so with shipments that were roughly 34 times smaller than AgiBot's total. This is not a market where the US is winning the volume war.

The pricing picture compounds the picture further. Unitree's G1 humanoid sells for $16,000 , a price point that has stunned Western robotics engineers who assumed cost parity was decades away. Tesla's Optimus is targeting $20,000 $30,000 for a planned run of 50,000 units at its Fremont facility in 2026, requiring conversion of legacy Model S and Model X production lines. The NEO humanoid from 1X Technologies, priced at $20,000, already had its entire 10,000-unit 2026 capacity booked before January, with a goal of 100,000 units annually by 2027. Boston Dynamics' electric Atlas, by contrast, is expected to carry a price tag of $140,000 $150,000 , a 9x premium over Unitree's current offering. China's AI2 Robotics has achieved a valuation of 20 billion yuan ($2.93 billion) while scaling manufacturing operations, not platform promises.

Why This Matters More Than People Think

The Omdia rankings capture hardware shipments , but the investor divergence is where the deeper story lives. US humanoid startups are being priced as artificial intelligence platforms, not manufacturing companies. Figure AI commands a valuation of at least $39 billion. Texas-based Apptronik reached a $5 billion valuation in February 2026. These multiples are justified by investors who believe that whoever wins the software and AI stack will capture the long-term value, even if Chinese firms dominate physical unit production. That thesis may be correct , but it is also the same thesis Western EV investors held about software-defined vehicles, and Chinese manufacturers did not stop at hardware.

The Omdia projection of 2.6 million units annually by 2035 , a 200x increase from 2025's baseline , is the number that should be focusing every strategic conversation. Whoever captures the manufacturing curve in the next two to three years will own a massive installed base before any AI premium becomes relevant. Chinese firms are not winning on AI benchmarks. They are winning on procurement decisions, deployment velocity, and the kind of industrial reliability that manufacturing scale produces over thousands of operational hours. Japan Airlines recently deployed humanoid robots at Haneda Airport. Siemens has units in its Erlangen factory. These are revenue-generating deployments, not demos. The customers choosing $16,000 Unitree robots over $140,000 Atlas units for warehouse logistics applications are not making a mistake , they are making a rational economic decision.

The Competitive Landscape

Boston Dynamics has spent three decades building the world's most recognized humanoid robot brand and only now is targeting commercial Atlas deployment in the 2026 2028 window. Figure AI raised at a $39 billion valuation on the promise of agentic AI embedded in physical form , a compelling story that requires both hardware at scale and software that outperforms whatever Chinese firms can build. Tesla's Optimus program is the wildcard: Elon Musk has committed to manufacturing one million humanoid robots annually at Fremont once the conversion is complete. If Optimus reaches even 25,000 units shipped in 2026, it signals that US manufacturing-scale robotics is viable. If it misses significantly, the investor case for US firms as AI-platform-plus-manufacturer collapses toward AI-platform-only , a much weaker competitive position.

On the Chinese side, AgiBot, Unitree, UBTech, LEJU Robotics, and Galbot , which raised $663 million , are not competing on AI benchmarks. They are competing on price, deployment velocity, and industrial reliability. The Chinese government's active support for domestic humanoid robotics mirrors its approach to EVs: subsidized manufacturing, preferential procurement, and a domestic market that provides the volume needed to drive cost curves down before export competition begins. The CNBC observation that "China is running the EV playbook on humanoid robots" is not an analogy , it is a description of explicit industrial policy.

Hidden Insight: The Installed Base Will Win

Here is the insight that the US robotics investment community is not yet grappling with honestly: the EV parallel is not just about manufacturing scale , it is about the point at which cost curves invert the software advantage. In EVs, Western brands assumed that their software ecosystems, brand recognition, and premium positioning would protect them even as Chinese competitors ramped volume. That assumption held until the cost gap became so wide that Chinese EVs began entering European and Asian markets at prices that no amount of premium positioning could justify. The same inflection point is approaching in humanoid robotics, and it will arrive faster than most US investors expect.

Consider the math plainly: if Unitree's G1 costs $16,000 and delivers 80% of the functional performance of a $140,000 Atlas for industrial applications, the remaining 20% of performance is only worth paying for in highly specialized contexts. Most factory, warehouse, and logistics deployments do not require that 20%. China does not need to win the AI benchmark war to win the commercial robotics market , it needs to win the procurement decisions of the 50,000 manufacturing plants globally deciding whether to automate the next shift. Those decisions are being made on price, reliability, and availability of trained service technicians, not on benchmark scores.

The most uncomfortable truth for US robotics investors is that they may be funding the wrong layer of the stack. The assumption is that AI software is where value concentrates , that the physical robot is just a commodity shell and the intelligence is the moat. But physical AI is different from cloud AI in one critical way: the hardware is the distribution channel. A humanoid robot embedded in a factory is a continuous data-collection asset and a continuously improving AI system. Whoever owns the installed base owns the training data. Whoever owns the training data builds the better model for that specific deployment environment. Whoever builds the better model improves the hardware. China does not need to start with the better AI. It just needs to get the hardware in the door first , and by the Omdia rankings, that is exactly what it is doing.

What to Watch Next

Watch AgiBot's 2026 shipment numbers, which Omdia will release in early 2027. If the company maintains or grows its share beyond the 13,000-unit 2025 baseline, the structural divergence becomes very difficult to reverse. Also watch Tesla's Fremont conversion timeline , if Optimus production hits even 25,000 units by Q4 2026, it signals that US manufacturing-scale robotics is viable and changes the competitive calculus. If it misses significantly, the gap between Chinese volume and US platform ambition becomes the defining story of the 2026 robotics cycle.

Track Unitree's price trajectory over the next 24 months. In EVs, the price compression from Chinese manufacturers came in waves: market entry at a slight discount, then aggressive cuts as scale allowed, then near-commodity pricing that forced global market restructuring. If Unitree or AgiBot announce a sub-$10,000 industrial humanoid by Q4 2026 , which their current scale trajectory makes feasible , it would dramatically accelerate enterprise procurement cycles and render much of the US premium positioning strategy obsolete. Also watch EU and Japanese industrial policy responses: the EV precedent taught Western governments to act earlier on manufacturing advantage gaps, and humanoid robotics is clearly on their radar as the next strategic sector requiring a policy response.

China does not need to build a better robot than America , it just needs to build more of them, cheaper, and get them in the door before American firms have finished debating the valuation.

---