China Takes the Top Five Humanoid Robot Patent Spots

Chinese startups occupy the first five places in a new ranking of humanoid-robot patent strength. They take six of the top ten spots overall. That is a serious result, but it is not a league table for the best robot you can buy.

The list comes from LexisNexis Intellectual Property Solutions, which analysed more than 26,000 patent families. Its Patent Asset Index rewards both portfolio size and patent quality. It does not score deployments, uptime, task success, price or customer support.

The top ten humanoid robotics startups by patent strength

The concentration at the top is the notable part. Fourier ranks first, while the Fourier GR-1 is already one of the better-documented Chinese humanoids in RoboZaps’ catalog. AgiBot, LimX, Pudu and Unitree complete an all-Chinese top five.

What the ranking actually measures

LexisNexis grouped the patents into three technical fields: interaction systems, morphology, and control and planning architectures. The Patent Asset Index then combines portfolio size with Technology Relevance and Market Coverage, also described as Competitive Impact. In plain English, it asks how much relevant intellectual property a company owns and how strong that property appears to be.

The country-level numbers are even more lopsided. At the end of 2025, China held 73% of active morphology-related patent families and 63% of global portfolio strength. The US held just 5% of those families but 11% of portfolio strength, with the highest average patent quality.

This helps explain why a fast-moving company such as AgiBot and its A2 humanoid can rank so highly while the US still retains a quality advantage per patent. China is producing far more relevant patent families; US portfolios carry more weight on average.

There is one error worth flagging on the source page. A summary line says seven of the ten startups are Chinese. The published table and the accompanying detailed paragraph contain six: Fourier, AgiBot, LimX, Pudu, Unitree and Leju. RoboZaps uses the table result: six.

What the list leaves out

This is explicitly a startup ranking. Public companies and subsidiaries of public companies are outside it, which excludes Tesla, the publicly listed UBTECH and Hyundai-owned Boston Dynamics. LexisNexis separately notes that UBTECH holds about 370 relevant patent families and devotes more than a quarter of its portfolio to the three core humanoid fields.

The ranking also says nothing directly about whether a robot can complete a paid shift. A strong patent portfolio may protect important technology or improve a company’s licensing position. It cannot tell a buyer whether the machine will meet a cycle-time target, recover from faults, integrate with existing software or receive competent local service.

Patent strength is meeting manufacturing scale

The patent result is not happening in isolation. TrendForce projects that Unitree and AgiBot will account for nearly 80% of global humanoid shipments in 2026. The Associated Press, citing Barclays and Omdia, reported that Chinese companies supplied about 85% of global humanoid shipments in 2025, with more than 13,000 units shipped and AgiBot and Unitree above 5,000 each.

That makes the Unitree G1 more than a viral demo machine: it sits inside a company with both patent depth and rising production volume. Still, shipment counts can include development units, research platforms and robots not yet delivering a repeatable commercial return.

A five-signal scorecard for buyers

US companies remain competitive on several of those other signals. Figure is sixth in the patent ranking, and the Figure 03 is being developed around factory and household use cases rather than patent counts alone. Apptronik and Agility Robotics also make the top ten, giving US buyers three domestic startups in the list.

The RoboZaps read

China’s lead is real. Taking the top five startup positions, controlling most active morphology patent families and building most of the world’s current humanoid volume is more than a publicity cycle. It points to a deep link between research, supply chains and manufacturing.

The US quality lead is real too. Its smaller patent pool carries disproportionate strength, and patent rankings do not capture the deployment evidence buyers ultimately need. The sensible conclusion is not that one country has already won. It is that China currently leads on patent scale and production, while individual purchasing decisions still have to be made robot by robot and task by task.