You Cannot Buy a Single Share of a Company That Earns Money Selling Humanoid Robots. The ETF Built on That Bet Is Up 66%.

There is no publicly traded company earning meaningful revenue from selling humanoid robots. This is the central claim underlying the investment case for KraneShares KOID ETF, and it is repeated in some form across every article written about the space. It is also, as far as anyone has been willing to show, completely unverified.

The ETF launched on Nasdaq on June 4, 2025, the first dedicated humanoid robotics index fund. In less than a year it returned 66.8% against the S&P 500 at 29.1% over the same period. As of late May 2026 it held $241 million in assets across 50 equal-weighted holdings, a mix of semiconductors, actuator manufacturers, and sensor companies, tracked against the MerQube Global Humanoid and Embodied Intelligence Index. The Morgan Stanley projection cited on KraneShares own website forecasts $5 trillion in annual humanoid revenue by 2050 with a 54% compound annual growth rate. Those are extraordinary numbers. They are also projections about 25 years out, built on assumptions about adoption that have yet to survive contact with a revenue line item.

The gap between what the ETF holds and what it claims to hold is worth sitting with. KOID is not a collection of robot makers. Its top holdings include Credo Technology, Infineon, STMicroelectronics, Harmonic Drive Systems, Renesas, and NXP Semiconductors: chip companies, component manufacturers, industrial automation suppliers. These are legitimate businesses. Some of them will almost certainly benefit as humanoid robotics scales. None of them has disclosed humanoid robot revenue as a meaningful standalone line. The index methodology selects companies based on a Humanoid and Embodied Intelligence Exposure Score, a proprietary ranking from MerQube, with the top 50 chosen by that score and market cap. That score is not the same as revenue.

The most cited justification for the category is the price trajectory. Humanoid robot average selling prices fell from roughly $85,000 in 2023 to around $25,000 in 2025, per KraneShares own analysis. That compression is real, and it matters for deployment economics. A robot that costs $25,000 and performs one full-shift equivalent of labor in a structured environment is in a different economic category than one priced at $85,000. The math becomes plausible for certain industrial applications, particularly in markets with acute labor shortages like Japanese manufacturing. The 54% CAGR Morgan Stanley projects would require that math to hold at scale, repeatedly, across a deployment environment that remains largely unproven.

The question nobody in this conversation is asking is whether the shape itself is right.

This matters because the commercial record is not encouraging. The most economically significant automation deployed at scale in the past two decades has resolutely refused to be humanoid. Amazon Robotics acquired Kiva Systems in 2012 for $775 million; Kivas warehouse bots are shelves on wheels. Intuitive Surgical has deployed 10,488 da Vinci surgical systems globally, per the company's Q2 2025 earnings. DJI owns roughly 70% of the global commercial drone market. None of these systems resembles a human. None needed to. They solved a specific problem at a specific cost point and found customers willing to pay for it.

The humanoid form factor has a different track record. Honda spent decades developing Asimo, which never sold a single commercial unit. Boston Dynamics spent years as a DARPA-funded humanoid project before pivoting to the quadrupedal Spot and later being sold to Hyundai. The commercial deployments that exist today of humanoid robots are pilot programs: Figure AI running limited trials with BMW, Tesla Optimus folding shirts in a lab, Agility Robotics Digit working in Amazon facilities under close human supervision. None of these represents a revenue line a financial analyst can point to.

This does not mean the form factor is wrong. A bipedal robot navigating a human-built environment, stairs and ladders and cluttered spaces, has genuine advantages over wheeled or quadrupedal alternatives. Tesla is serious about Optimus. Figure has real customers. Unitree is growing. The price compression is real and it changes what is possible. But the investment thesis and the engineering thesis are not the same thing. Capital markets have pre-priced a decade of commercial deployment into an ETF that holds semiconductor suppliers and bearings manufacturers, before any of those companies can demonstrate that humanoid-specific revenue exists at all.

There is real money in this trade, potentially. There is also real risk that the shape everyone is building toward is the one their investors expect, rather than the one the deployment economics actually demand. The ETF is up 66%. The robots are not yet.