The new battleground
In humanoid robotics, hardware is no longer the main battleground: according to The Korea Herald, competition now centers on the ability to secure the ecosystem of so-called physical intelligence—that is, the set of components, models, and data that make a robot operational. This premise gives rise to two opposing strategies, which the South Korean daily describes as a decisive test ahead of commercialization.
China's bet on vertical integration
Chinese players in the sector are investing venture capital in local startups developing AI semiconductors, precision sensors, robotic hands, and physical-intelligence models, with the goal of building a tightly integrated domestic supply chain that is less exposed to foreign constraints. According to media reports on Wednesday cited by the source, AgiBot has expanded its incubation program, aiming to support more than 50 new companies over three years after already funding more than 20 startups in the past year.
Unitree is pursuing a similar logic, with a joint program with Sequoia China to fund developers of components and physical-intelligence solutions and to build a network of more than 200 partners, with collaboration commitments estimated at several billion yuan over the next five years. The source attributes this choice to domestic price pressure and U.S. chip restrictions. Arnold Platz, founder of the Hungarian company General Future Holding, argues that in a price war, owning the entire supply chain is the condition for defending margins; Patrick van Stokhem, of Breman Installatiegroep, describes the model as a network of minority shareholdings designed to control the scarcest components, from hands to tactile sensors to training data.
Hyundai's horizontal path and its weaknesses
Boston Dynamics, a subsidiary of Hyundai Motor Group, has chosen the opposite path: it uses the group's automotive manufacturing and industrial plants as a testing ground, but entrusts artificial-intelligence and vision-language-action models to Google DeepMind and digital simulation to Nvidia. According to the source, this is an immediate shortcut, but its cost is operational control and development speed.
Platz warns that depending on a partner's brain means moving at the partner's pace, with a roadmap shaped by others' priorities, and cites the recent leadership change at Nvidia as an example of how quickly an ally can become an unknown. Van Stokhem adds the risk of slow feedback loops compared with Chinese competitors, which can update hardware and software within weeks, and the risk that Hyundai's factory data will end up training models that partners could license to competitors.
An outcome still open
The experts cited remain divided. Platz predicts that the gap in actuators, where China is still behind, will be measured in months or a few years, not decades, and that Chinese manufacturers may then lead on almost every component in terms of price-performance. Yun Dong-won, a professor at the Daegu Gyeongbuk Institute of Science and Technology, sees Hyundai's approach as the Korean "fast follower" model and warns that without investment in proprietary technologies the group risks falling behind over time. Van Stokhem, by contrast, believes that Western alliances have an advantage in regulated industrial facilities in trusted markets, where safety and compliance are decisive, and that the winner will be decided on the factory floor, among operational continuity, certifications, and integration. It remains to be verified how substantial the announced financial commitments actually are and whether the two strategies hold up under the test of commercialization.
