Q&A Series: Humanoid robots could be AI’s next big wave

Artificial intelligence is increasingly moving beyond the digital world and into the physical environment. Advances in AI models, machine learning, and robotics hardware are enabling machines to perceive their surroundings, make decisions and perform tasks in the real world. Humanoid robots are emerging as one of the most prominent applications of this trend, with potential use cases spanning manufacturing, logistics, healthcare and other labour-intensive industries.

As the technology advances and real-world deployments increase, investors are also beginning to explore the potential of humanoid robotics as a long-term growth theme.

In this article, we share insights from KraneShares on how humanoid robots differ from traditional industrial robots, the investment opportunities across the value chain, and whether it is too early to invest in the theme.

Introducing the KraneShares Global Humanoid Robotics and Physical AI Index ETF

The KraneShares Global Humanoid Robotics and Physical AI Index ETF (NASDAQ: KOID) is the first US-listed ETF focused on humanoid robotics. It tracks the MerQube Global Humanoid Robotics and Physical AI Index, providing broad exposure to the humanoid robotics ecosystem, from companies that design and manufacture humanoid robots to the supply chains providing key components and technologies that enable them.

Table 1: Key information about the ETF

ETF Details
Underlying Index	MerQube Global Humanoid Robotics and Physical AI Index
Base Currency	USD
Trading Currency	USD
Nasdaq Ticker	KOID
Listing Date	4 June 2025
Number of Holdings	50
Assets Under Management	USD 338,203,074
Trading Board Lot Size	1 unit
Expense Ratio (Net)	0.69
Dividend Distribution Frequency	Annually
Source: KraneShares. Data as of 21 Sep 2026

1. What is physical AI, what is embodied AI, and what is a humanoid robot? How do these three terms relate to one another?

Think of it as a progression. For years, AI lived entirely in the digital world—answering questions, generating images, writing code, summarising information on a screen. Then we moved toward AI agents that can autonomously complete digital tasks. Physical AI is the next step: giving AI the ability to perceive, reason about, and act in the physical world through robots, autonomous vehicles, drones, and other machines.

Embodied AI is essentially AI with a body. You take the intelligence—the brain—and connect it to sensors, actuators, and a physical system that can interact with the real world. That embodiment does not have to look human. It could be a robotic arm, a warehouse robot, or a self-driving vehicle.

A humanoid robot is one specific form of embodied AI, where the body is designed around the human form—typically with arms, hands, a torso, and legs.

The important point from an investment perspective is that physical AI potentially expands the economic impact of all the capital we have already deployed into AI infrastructure beyond the digital economy. AI moves from answering questions and writing code to doing physical work—manufacturing products, moving goods, assembling components, inspecting infrastructure, or performing tasks in environments built for the human body. That transition from digital to physical is what makes this an investable theme, not just a technology trend.

2. What are the real-world use cases for humanoids today, and is the human form factor really the most efficient way to perform these tasks?

The use cases are expanding quickly. In manufacturing, humanoids are performing quality inspection, parts handling, and assembly tasks alongside human workers. In logistics, they are sorting packages, loading pallets, and navigating warehouse floors. Unitree shipped over 5,500 humanoid robots in 2025—more than any other company globally—which tells you this is no longer a lab demonstration. Real customers are buying these machines, deploying them, and reordering.

We are also seeing early deployments in commercial services, retail environments, and hazardous inspection work where sending a human is either dangerous or impractical. China's government has designated humanoid robotics a national priority, and multiple provinces are running subsidised pilot programmes in factories, which is accelerating the feedback loop between deployment, data collection, and product improvement.

Now, is the human form always the most efficient design for a given task? Not necessarily. A fixed robotic arm on an automotive production line will always outperform a humanoid at a single repetitive motion. But that is not the right comparison.

Think of it this way: a forklift is the most efficient way to move a pallet across a flat warehouse floor. But a forklift cannot climb stairs, open a door, pick an irregular item off a shelf, restock a display, and then walk to a different floor to do something entirely different. A humanoid can potentially do all of those things on a single platform—because our entire built environment was designed around the human body. The height of a countertop, the width of a doorway, the layout of a factory workstation, the spacing of warehouse aisles—all of it assumes a human operator. A humanoid fits into that world without requiring the infrastructure to be rebuilt around it.

That generality is what makes the economics compelling. Instead of deploying ten specialised machines for ten different tasks, you deploy one adaptable platform that can be retrained through software. The cost of retraining a humanoid for a new task is a software update; the cost of retooling a specialised machine is a capital expenditure. As AI models improve and training costs fall, that flexibility advantage compounds.

There is also a natural spectrum within the form factor. Some applications only require two arms on a fixed base. Others need a wheeled mobile torso. Full bipedal locomotion—legs—adds cost and complexity, but it unlocks environments like construction sites, disaster zones, and multi-storey facilities where wheels simply cannot go. The market will support different configurations at different price points, and KOID is designed to capture value across that entire spectrum—not just the humanoid integrators, but the sensors, actuators, semiconductors, and materials that every configuration requires.

3. Is it too early to invest in humanoids, given how early we still are in terms of adoption and commercialisation? Why invest in humanoids specifically rather than industrial and warehouse robotics more broadly?

That is the question we hear most often, and the honest answer is that yes, full-scale commercial adoption of humanoid robots is still in its early stages. Most deployments today are pilots. But from an investment standpoint, that is precisely the point.

If you wait until humanoids are fully commercialised and the revenue numbers are obvious, you will be investing after the market has already priced in most of the upside. The analogy I often use is early-stage AI infrastructure investment. When NVIDIA was primarily selling GPUs for gaming and data centres had not yet been rewired for AI workloads, the market underappreciated how large the opportunity could be. The investors who did well were the ones who understood the direction of travel before unit economics were fully proven.

What we are seeing today in humanoids is a similar inflection. Morgan Stanley projects the global humanoid industry could generate USD 5 trillion in annual revenue by 2050 and forecasts a 54 percent compound annual growth rate over the next decade. Unitree shipped over 5,500 humanoid robots in 2025 and posted 60 percent gross margins—proof that scalable profitability in humanoid hardware is achievable, not theoretical.

As for why humanoids rather than industrial robotics broadly, the difference is the addressable market. Traditional industrial automation targets repetitive, structured tasks in controlled environments. The installed base is mature, growth is incremental, and the value chain is well understood and well priced. Humanoids address a fundamentally different opportunity: unstructured environments, variable tasks, and the vast categories of physical work that automation has not yet reached. The Morgan Stanley estimate is that the serviceable addressable market for humanoids extends into manufacturing, logistics, elder care, agriculture, and household services—sectors where a flexible, general-purpose form factor is the enabling condition, not just an engineering preference.

KOID gives you exposure to that broader opportunity—not just the robot OEMs, but the full supply chain—while a pure industrial robotics fund concentrates you in the portion of the market that is already well penetrated.

4. Could you walk us through the humanoid value chain and where KOID’s holdings sit across it?

We generally think about the humanoid value chain as the brain, the body and the integrator.

The Brain layer is the intelligence stack—the semiconductors, sensors, and computing platforms that enable a robot to perceive, reason, and act. Holdings here include NVIDIA, which provides the GPU compute for training and inference; Horizon Robotics, whose edge AI processors power autonomous decision-making; Texas Instruments and Analog Devices, which supply the mixed-signal and embedded processing chips that sit at the sensor-processor interface; NXP Semiconductors and Infineon, which provide automotive-grade microcontrollers increasingly adapted for robotic control systems.

The Body layer covers the mechanical and physical systems—actuation, sensing, materials, and motion. Harmonic Drive Systems and Leader Harmonious Drive supply the precision reduction gears that translate motor output into controlled movement at every robot joint. Nabtesco provides complementary reduction gear technology. Hiwin Technologies and THK supply the linear motion systems and ball screws that enable smooth, precise movement. Sensata Technologies and Melexis provide sensors—force, torque, and position—that give the robot proprioceptive and environmental awareness. Cognex handles machine vision. MP Materials and Lynas Rare Earths supply the rare earth elements critical for the permanent magnets inside the electric motors that actuate every limb. China Northern Rare Earth sits in the same supply chain. Schaeffler and RBC Bearings provide the precision bearings that support rotational components.

The Integrators are the companies that bring Brain and Body together into complete robotic systems. Tesla, through its Optimus programme, is building a vertically integrated humanoid from proprietary motors to its own AI software stack. UBTech Robotics, listed in Hong Kong, is one of the few publicly traded pure-play humanoid OEMs, with robots deployed across education, commercial, and industrial settings. Unitree Robotics, which just listed on Shanghai's STAR Market, is the world's leading humanoid shipper by unit volume. Rainbow Robotics in Korea is another dedicated humanoid manufacturer. Doosan in Korea brings its industrial robotics heritage to collaborative and humanoid platforms. Symbotic integrates warehouse automation with AI-driven robotic systems. Jabil, Magna International, and Aptiv contribute contract manufacturing and system integration capabilities.

The equal-weight methodology means KOID does not let any single layer or company dominate the portfolio, which we think is appropriate at this stage of the market.

5. Which country is currently leading in humanoid robotics, and does KOID’s geographic allocation reflect that?

This is a genuinely multipolar race, and the geographic allocation of KOID reflects that reality.

The United States leads in the software and compute stack. NVIDIA's training and inference platforms are foundational, Tesla's Optimus programme is one of the most vertically integrated efforts in the world, and the US semiconductor ecosystem—Analog Devices, Texas Instruments, Monolithic Power Systems—supplies critical components across the global supply chain. The US is also home to companies like Boston Dynamics, though as a Hyundai subsidiary it is not separately listed.

China is the clear leader in manufacturing scale and cost competitiveness, and arguably the leader in commercialisation speed. Unitree shipped over 5,500 humanoids in 2025—more than any other company globally. UBTech has been publicly traded since December 2023 and has the broadest product line of any listed humanoid OEM. China also dominates in precision components: Leader Harmonious Drive, Zhejiang Shuanghuan, and Jiangsu Hengli are all leaders in reduction gears and hydraulic systems for robotic actuation. China's government has made humanoid robotics an explicit national priority, which is driving capital formation and adoption at a pace the private sector alone would not support.

Japan has deep structural advantages in precision manufacturing—Harmonic Drive Systems, Nabtesco, Nidec, Keyence, and THK are all Japanese companies with decades of leadership in the motion control and sensing components that are essential to any humanoid design.

South Korea, through Rainbow Robotics and Doosan, is competitive in humanoid integration, and benefits from strong government support for robotics as part of its industrial policy.

Europe contributes through the semiconductor and sensor supply chain—Infineon, STMicroelectronics, NXP, Melexis, Hexagon, and Schaeffler are all European companies with critical positions in the humanoid ecosystem.

KOID's allocation reflects this landscape. The United States, China, and Japan collectively represent the largest country exposures, with meaningful positions in South Korea, Europe, and Taiwan. No single country dominates, which mirrors the reality that the humanoid supply chain is inherently global—no one country controls every critical node.

6. Unitree recently had a blockbuster IPO on Shanghai’s STAR Market. Could the company eventually become eligible for inclusion in KOID?

KOID added Unitree Robotics following its STAR Market listing in August 2026. It is following the index IPO fast-track methodology and through the Qualified Foreign Institutional Investor (QFII) program since it is not stock connect eligible yet. As of September 18, 2026, Unitree is held at approximately 1.5 percent of net assets.

The Unitree IPO was a landmark event for the humanoid robotics sector. The company priced at an implied valuation of roughly USD 9 billion and surged over 460 percent on its first trading day. Unitree shipped more than 5,500 humanoid robots in 2025, making it the world's leading humanoid manufacturer by unit volume, and posted 2025 revenue of RMB 1.7 billion with gross margins above 60 percent. DeepSeek participated as a strategic investor in the offering.

Because KOID uses an equal-weight methodology, Unitree's position size is comparable to other constituents rather than being oversized by its IPO-day market capitalisation—which we view as appropriate given the stock's early-stage volatility.

What makes Unitree particularly significant for investors is that it validates the commercial viability of humanoid hardware as a category. It is one of very few companies anywhere in the world producing humanoid robots at scale, with real revenue, real margins, and a credible path to profitability—not just a demonstration programme or a corporate side project.

7. How does KOID compare with other humanoid-focused ETFs such as BOTT and HUMN, and could its equal-weighting approach potentially limit the ETF’s upside?

There are meaningful structural differences between the three funds, and we think those differences matter more than investors typically appreciate.

BOTT (Themes Humanoid Robotics ETF) tracks the Solactive Global Humanoid Robotics Index. It is significantly more concentrated. BOTT's concentration tilts heavily toward Chinese component makers and Tesla, which means its performance is driven disproportionately by a handful of names. With approximately USD 60–67 million in AUM, BOTT is meaningfully smaller than KOID, which can lead to wider bid-ask spreads and less liquidity for institutional allocators.

HUMN (Roundhill Humanoid Robotics ETF) is actively managed. HUMN holds around 47 names with a 75-basis-point expense ratio. The active mandate gives Roundhill the flexibility to overweight or underweight names based on conviction, but it also introduces manager selection risk and the question of whether an active approach adds value in a nascent market where the winners are not yet obvious.

KOID uses an equal-weight methodology across approximately 50 holdings, rebalanced quarterly. With approximately USD 338 million in AUM, KOID is the largest of the three funds, which provides better trading liquidity. At this stage of the cycle, we believe diversified, equal-weight exposure to the entire humanoid value chain is a more prudent approach than making concentrated bets on individual names—and the AUM and liquidity profile of KOID supports institutional-quality execution. Below are the annualized return and risk for these ETFs. Since inception, KOID offered the highest return with lowest volatility.

6/30/2025-8/31/2026	KOID	HUMN	BOTT
Annualized Return	34.9%	18.2%	33.3%
Annualized Risk	33.6%	37.4%	49.7%

Data from Bloomberg as of 8/31/2026. The performance data quoted represents past performance. The investment return and principal value of an investment will fluctuate so that an investor’s shares, when redeemed or sold, may be worth more or less than the original cost. Current performance may be lower or higher than the performance quoted. For performance data current to the last month-end, please visit kraneshares.com/koid.