Unitree Robotics will be included in the world's largest humanoid robot ETF, but robots still need to move beyond the performance arena.

Unitree Robotics will be included in KraneShares' KOID (Koid, the world's largest humanoid robot ETF) and KSTR (which tracks the STAR Market). KOID covers the entire industry chain, with a significant proportion of Chinese component manufacturers. KraneShares does not set a limit on its holdings in China. The industry currently faces a dual challenge: commercialization is constrained by high costs and difficulties in implementing solutions for non-standard scenarios; while the hardware technology is mature, shortcomings such as the weak generalization ability of the AI "brain" and poor adaptability to complex environments urgently need to be addressed.

Stepping into KraneShares' New York office, known as the "pioneer of Chinese concept stock ETFs," several Unitree humanoid robots are dancing. Just days ago, Unitree Technology was listed on the STAR Market; now, this American asset management firm is preparing to bring Unitree into the ETF accounts of overseas investors.

Following Unitree Robotics' listing on the STAR Market, Tencent Finance has exclusively learned that Unitree will be included in two KraneShares ETFs: KSTR, which is dual-listed on the NYSE and LSE and tracks the STAR Market, and KOID, listed on Nasdaq. The latter is the first and largest humanoid robot-themed ETF in the United States and globally.

In an interview with Tencent Finance, KraneShares CEO Jonathan Krane stated that Morgan Stanley predicts over 1 billion robots will be in use globally over the next 25 years, representing a market size of $5 trillion. However, he also admitted that the challenges have only just begun, with most achievements still focused on entertainment performances, and core problems such as algorithm generalization remaining unsolved.

01 Unitree Robotics will be included in the largest humanoid robot ETF

KOID has undergone more than a year of refinement from design to launch, and has already attracted a continuous inflow of global institutional and individual investors.

Jonathan Krane stated that Unitree's IPO is a milestone, "showing the world the growth rate of this industry and allowing global investors to see the investment opportunities in this sector more clearly."

It's worth noting that KOID doesn't bet on a single OEM, but rather covers the entire humanoid robot industry chain. The fund tracks the MerQube Global Humanoid Robotics and Physical AI Index, holding 51 to 58 global stocks, with technology (semiconductors and electronics) accounting for approximately 45%, industrial manufacturing approximately 33%, and consumer goods approximately 13%; regionally, North America and Asia each account for approximately 41%, and Europe approximately 15%.

Global leaders are present in every key link of the industry chain: in rare earths, Lynas of Australia is the largest rare earth producer outside of China; in precision transmission, Harmonic Drive of Japan and Green Harmonic Drive cover two technical routes; in semiconductors, STMicroelectronics, Infineon, and Texas Instruments span motor control and sensor signal chains; in connectors, TE Connectivity supports the "nerves and skeleton" of robots; in vision perception, Hexagon of Sweden and Cognex of the United States each have their own layouts; and in complete machines, Horizon Robotics, Rainbow Robotics of South Korea (with Samsung holding shares), and Symbotic, a leader in warehouse automation, are all included.

The presence of China's supply chain is particularly prominent. Companies like Kewell (servo drives), Lingyi iTech (precision structural components), and Keli Sensing, all listed on the A-share market, have already entered their holdings. Jonathan Krane explicitly stated that he has no upper limit on Chinese companies—"Chinese companies are already at the forefront of the world in core components."

02. The US and China may still cooperate amidst competition.

Geopolitical uncertainties have not deterred Jonathan Krane from investing in Chinese assets. In his view, China is already at the forefront of the world in core components such as actuators, servo motors, and batteries, while the United States' strengths lie in software and AI; the two are naturally complementary, and industrial collaboration is an inevitable trend.

In his view, Nvidia chips are used by both Unitree Robotics and American robotics companies. "The flow of cutting-edge technology across geographical boundaries is itself an unbreakable bond for Sino-US cooperation."

Several experts pointed out to Tencent Finance that the robotics industries in China and the United States have different focuses: North American companies like Figure and Tesla are more focused on the ultimate goal of combining bipedal forms with large models, and their R&D and verification cycles are relatively long; Chinese companies, with their complete supply chains and rapid delivery capabilities, are showing a "hundred flowers blooming" trend and are better at differentiating their products based on their own technological advantages—companies that are good at control demonstrate multiple forms, while those that focus on the "brain" use wheeled chassis to avoid complex motion control.

Regarding the recent potential restrictions imposed by the U.S. Federal Communications Commission (FCC) on optical modules and robots, Jonathan Krane's assessment is that commercial logic will ultimately prevail. There are precedents for this – in the electric vehicle supply chain, while vehicle manufacturers struggle to enter North America, battery companies like Guoxuan High-Tech have already established factories in the U.S., and CATL plays a crucial role. The same pattern can be observed in the robot supply chain.

03 Embodied intelligence still faces challenges in generalization and practical application.

It is undeniable that Yushu's humanoid robots have become a hit on social media at home and abroad, but most of the achievements are still limited to performances and demonstrations, with limited commercial applications.

Several industry insiders told Tencent Finance that in terms of commercialization, the initial capital investment is high and the pressure for ROI is significant. Non-standardized scenarios—hospitals, hotels, and homes—present far greater technical challenges than warehousing and factories. The industry must first successfully implement structured scenarios before discussing scale; the unit price of the entire machine also needs to be further reduced to truly achieve widespread adoption in both B2B and B2C markets.

Technically, the hardware is relatively mature, but the weakness lies in the "brain." While Unitree's WVLA 2.0, released in May 2026, integrates a world model and VLA architecture, industry insiders admit it still has significant shortcomings: a lack of quantitative evaluation of continuous task success rates; limited generalization capabilities, currently mostly demonstrated in controlled conference room environments, struggling with complex situations such as varying table heights, lighting, and clutter; insufficient memory and anomaly handling for long-duration tasks; delays in Action Chunking mode during sudden environmental changes, leading to jittery movements; and room for improvement in the precision of grasping special objects such as transparent, reflective, and slender objects.

Market competition is equally fierce. Hundreds of startups are vying for dominance, the survival rate of OEMs is questionable, and overseas venture capitalists generally still prefer mature industrial automation sectors, making the early-stage funding environment for humanoid robots relatively cautious.

Jonathan Krane believes that commercial applications in factories, hotels, and hospitals will gradually materialize, and humanoid robots will eventually enter ordinary households. "Efficiency improvements and practical value will soon permeate society and the labor market." However, this path will take longer than the current enthusiasm in the capital markets. Unitree's IPO is a starting point, not an end.