In early 2025, the United States Commerce Department added several Chinese robotics and AI companies to its Entity List — a roster of foreign organisations that American firms cannot supply with technology without a special licence. The specific companies named were not household names in Western robotics coverage. But the action signalled something that has been building for several years and that the humanoid robotics field has not yet fully reckoned with: governments have decided that advanced robotics is a strategic technology, and they are beginning to treat it accordingly.
Export controls, trade restrictions, and industrial policy are not typically covered alongside motor torque specs and manipulation benchmarks. They should be. The trajectory of humanoid robotics — which companies succeed, which supply chains are viable, which countries lead — will be shaped as much by what governments permit as by what engineers build. Understanding the policy layer is not optional context. For anyone trying to follow this field seriously, it is part of the story.
What Export Controls Actually Control
Export controls are legal restrictions on what technology can be sent to which destinations or entities. In the United States, they are administered primarily by two agencies: the Commerce Department's Bureau of Industry and Security (BIS), which controls dual-use items through the Export Administration Regulations (EAR), and the State Department's Directorate of Defense Trade Controls (DDTC), which handles items on the United States Munitions List (USML).
The EAR uses a classification system called Export Control Classification Numbers (ECCNs). Each ECCN describes a category of technology and specifies which countries require a licence for export and which are freely permitted. Technologies with military or intelligence applications, or those deemed to have significant national security implications, receive stricter classifications. Exporting a controlled item without the required licence is a federal crime.
Humanoid robots sit in an interesting position within this framework. A commercial humanoid robot sold for warehouse logistics does not, on its face, look like a munitions item. But humanoid robots contain — or depend on — several categories of technology that are already subject to significant export controls: advanced semiconductors, certain AI software and models, high-precision sensors, and, in some cases, motor control systems that have direct military analogues. The robot as a complete system may not be controlled, but its most strategically sensitive components often are.
The practical implication is that a US humanoid robot company selling its product internationally, or sourcing components globally, is operating within an export control environment whether or not it is thinking about that environment explicitly. The controls are not yet specifically tailored to humanoid robots as a category. But they cover enough of what goes into one that compliance is a genuine operational concern.
The Semiconductor Layer
The most consequential existing set of export controls for humanoid robotics is not specific to robotics at all. The Biden administration's October 2022 semiconductor export restrictions — significantly expanded in subsequent updates — placed sweeping limits on the export of advanced chips and chip-making equipment to China. The restrictions target the most capable AI accelerators produced by companies including Nvidia, as well as the equipment used to manufacture advanced chips, most critically the extreme ultraviolet (EUV) lithography machines made by the Dutch company ASML.
These restrictions matter for humanoid robotics because the most capable humanoid systems are computationally intensive. The AI inference that drives perception, motion planning, and task reasoning in a modern humanoid robot runs on chips. The most powerful chips available — and, increasingly, chips specifically designed for edge AI inference in mobile robots — are exactly the category the export controls are targeting. A Chinese company developing a competitive humanoid robot cannot legally obtain the most advanced US-origin AI accelerators. It must either develop its own chips (which China is investing heavily in, with mixed results so far), use less capable alternatives, or find other ways around the constraint.
The chip controls do not prevent Chinese companies from building humanoid robots. Unitree, which sells its G1 humanoid at a price point that undercuts Western competitors substantially, has demonstrated capable systems using available hardware. But the controls do create a real, if imprecise, performance ceiling on what Chinese companies can build with commercially available components — and that ceiling matters in a field where compute is a genuine constraint on capability.
China's Humanoid Programme Is Not a Coincidence
China's engagement in humanoid robotics is extensive, heavily state-directed, and explicitly framed in national strategy terms. The Ministry of Industry and Information Technology issued guidance in late 2023 designating humanoid robots as a priority sector for development, with a target of establishing a viable domestic humanoid industry by 2025 and global competitiveness by 2027. State-owned enterprises, provincial governments, and national investment funds have channelled substantial capital into the sector. The list of Chinese humanoid companies — Unitree, UBTECH, Fourier Intelligence, Agibot, Deeprobotics, and others — reflects a policy environment designed to produce multiple competing domestic players rather than a single national champion.
The strategic framing is explicit. Chinese officials and state media have described humanoid robots as a key technology for maintaining manufacturing competitiveness as labour costs rise and for reducing dependence on foreign automation technology. The parallel with the semiconductor sector — where Chinese strategic investment was accelerated significantly by the recognition that external supply could be constrained — is not subtle.
From a Western policy perspective, this framing creates a dilemma. The technologies that make humanoid robots capable — advanced AI models, high-precision sensors, powerful chips — are predominantly developed in the United States, Europe, and Japan. Selling those technologies freely to Chinese competitors accelerates a programme that is explicitly aimed at displacing Western companies from global markets. Restricting them too aggressively pushes Chinese development toward self-sufficiency and may disadvantage Western companies that want to sell in the Chinese market or source components there. Neither option is clean.
The Component Sourcing Problem
Western humanoid robot companies have their own supply chain vulnerabilities, and those vulnerabilities intersect with trade policy in ways that are not always clearly acknowledged.
Several of the most important physical components in a humanoid robot — harmonic drive gearboxes (which allow precise, high-torque motion in a compact form), high-density lithium-ion cells, and certain categories of rare earth magnets used in motors — are either predominantly manufactured in China or depend on Chinese-controlled materials at some point in their supply chain. Harmonic drives, for example, are produced by a relatively small number of manufacturers globally, with significant production capacity in Japan and China. Rare earth elements used in the powerful permanent magnets that drive many robot actuators are mined and processed overwhelmingly in China.
This creates a structural tension. American and European humanoid robot companies are developing products that compete with Chinese companies, in a policy environment that increasingly restricts technology flows to China — while simultaneously depending on Chinese-controlled supply chains for some of their own key inputs. The tension is not hypothetical: China has previously demonstrated willingness to restrict exports of rare earth materials for strategic reasons, as it did in 2010 in response to a territorial dispute with Japan, temporarily disrupting global supply chains for electronics and precision machinery.
The humanoid robotics industry has not yet had a supply chain crisis of that kind, but the structural conditions for one exist. Companies that have thought carefully about this are working on supply diversification — qualifying alternative gearbox suppliers, supporting domestic rare earth processing investment, redesigning actuator systems to reduce dependence on specific material categories. Companies that haven't thought carefully about it are carrying risks they may not have fully quantified.
The Investment Screening Layer
Beyond export controls on technology, a second policy mechanism is shaping humanoid robotics: restrictions on foreign investment in strategic technology sectors. In the United States, the Committee on Foreign Investment in the United States (CFIUS) reviews foreign acquisitions of, and certain investments in, American companies for national security implications. The scope of CFIUS review has expanded significantly since 2018, when the Foreign Investment Risk Review Modernization Act (FIRRMA) broadened its authority to cover minority investments in companies involved in critical technologies, critical infrastructure, and sensitive personal data.
Advanced robotics is explicitly listed as a critical technology category under FIRRMA. A Chinese investor taking a significant stake in an American humanoid robot company would face CFIUS review, and that review could result in conditions being placed on the investment or the transaction being blocked entirely. Several such reviews have occurred in adjacent technology sectors, with some resulting in forced divestiture of existing investments.
The practical effect is that Chinese capital, which flows freely into many technology sectors globally, faces significant constraints when directed at American robotics companies working at the frontier. This shapes funding structures: it pushes American robotics companies toward US-domiciled investors, limits certain partnership and joint venture structures with Chinese entities, and complicates commercial relationships that might otherwise be straightforward. It also, notably, does not apply symmetrically: American investment in Chinese robotics companies faces no equivalent CFIUS-style screen in China, though other regulatory mechanisms limit foreign ownership in strategic sectors there.
What This Means for the Field
The policy layer does not determine who wins in humanoid robotics. Engineering capability, manufacturing cost, reliability at scale, and the ability to actually solve deployment problems in real environments will matter more than trade policy in determining which companies are still operating a decade from now. But policy sets constraints within which the competition occurs — and those constraints are tightening.
For companies in the field, the immediate implications are practical: export compliance programmes, supply chain audits, investment structure decisions, and in some cases product architecture choices (which chips to design around, which components to source from where) are now genuinely important operational questions, not just legal department paperwork.
For the broader trajectory of the field, the more significant question is whether the increasing strategic framing of humanoid robotics by governments on both sides will accelerate parallel development — producing two distinct technology ecosystems with limited interoperability — or whether the commercial incentives for cross-border collaboration are strong enough to keep the field more integrated than current policy direction suggests. The semiconductor industry offers a cautionary comparison: a field that was deeply globally integrated a decade ago and is now, deliberately and with enormous cost, being restructured along national lines.
Humanoid robotics is earlier in that process. Whether it follows the same trajectory depends partly on decisions being made now, in policy offices and boardrooms, that rarely appear in the coverage of the latest demo video.