Letter 03 asked whether Databricks was the shovel-seller of the AI infrastructure gold rush. This letter asks the same question about a different gold rush — the humanoid robot — and the answer sits in a part of the machine almost nobody watching the demo videos ever looks at: the joint. A single mechanical component, the harmonic or cycloidal reduction gear, runs a large share of a humanoid's entire bill of materials. A handful of companies make it, most of them publicly investable today, and in May 2026 one of them signed a deal to supply over a million of them.
Watch a humanoid robot demo and the eye goes to the whole machine — the walk, the grip, the face. The money doesn't sit there. Inside every one of those joints is a precision reduction gear, either a harmonic drive or a cycloidal (RV) reducer, converting a motor's fast, weak spin into the slow, powerful, controlled motion an arm or a leg actually needs. It's an unglamorous part nobody puts in a keynote. It's also, by most estimates, one of the largest single line items in a humanoid's entire bill of materials — the actuator assembly built around it runs roughly half the cost of the machine. Letter 03 called this pattern the pick-and-shovel play. This is the same logic, one supply chain layer down from where the humanoid headlines usually stop.
The reducer market is genuinely concentrated. Japan's Harmonic Drive Systems (Tokyo: 6324, also traded over-the-counter in the US as HSYDF) is the dominant name in high-precision harmonic gears, and it's a real, actively-covered public company — analysts have been explicitly revising its investment case around humanoid robotics demand even as its legacy automotive customer base (tied to Nissan) works through its own troubles. Nabtesco (Tokyo: 6268) holds the equivalent position in cycloidal RV reducers, the technology of choice for heavier industrial robot joints, and sits inside the KraneShares Global Humanoid Robotics ETF (KOID) as one of its actuation-layer holdings. Both are Japanese incumbents with decades of precision-manufacturing history behind them — not startups betting on a thesis, but established suppliers a Schaeffler-scale customer already trusts.
China is scaling a cost-competitive alternative fast. Green Harmonic (Shanghai: 688017), also referred to as Leader Drive, makes the same class of strain-wave reducers and is reportedly a primary supplier into Tesla's Mexico operations — a real commercial relationship, not a speculative one. Its market cap has moved with the theme: up roughly 89% since KOID's inception. The pattern across this whole layer is consistent — Japan holds the precision high end, China is undercutting on cost and scaling output, and both are structurally more exposed to the humanoid cycle than almost any name actually building a robot chassis.
Nobody photographs the gear inside the ankle joint. Every humanoid on stage this year needed one anyway — and the company that made it didn't have to bet on which robot, or which country, would win.
On May 13, 2026, Schaeffler — the German automotive-parts incumbent, not a robotics startup — signed two agreements with UK humanoid maker Humanoid AI on the same day. First, a commitment to deploy 1,000 to 2,000 robots across Schaeffler's own manufacturing plants by 2032. Second, and more relevant to this letter's thesis, a five-year actuator supply agreement running through 2031, under which Schaeffler will supply more than half of Humanoid AI's wheeled-platform joint actuator demand — over one million units. That is a century-old auto-parts company converting its existing precision-manufacturing base directly into humanoid supply, at a scale that turns a thematic bet into a contracted, multi-year revenue line. The same week, Figure AI livestreamed its Helix-02 system autonomously sorting packages for over 80 hours with zero failures — the kind of commercial-viability proof point that makes contracts like Schaeffler's look less speculative in hindsight.
Two more names round out the "boring but essential" side of this supply chain. Zhejiang Sanhua Intelligent Controls, better known for automotive and HVAC thermal-management components, has become a real humanoid supplier because dense actuators and batteries generate heat that has to be precisely controlled — Sanhua's valves, heat exchangers, and thermal modules are a natural fit, and the stock has moved accordingly (roughly +96% since KOID's inception). Further upstream still sits the raw-material chokepoint this letter series has flagged before in other contexts: rare-earth magnets are essential to the high-torque motors inside every one of these actuators, and China controls the overwhelming majority of global processing capacity. Lynas Rare Earth, Australian-headquartered and mining cerium, praseodymium, and neodymium outside Chinese control, is one of the only publicly-investable alternatives to that concentration — and it has been the single best-performing name in KOID's basket, up roughly 131% since the ETF's inception.
Every name above is individually investable, but stock-picking across gear manufacturers, thermal specialists, and a rare-earth miner is a genuinely different research burden than buying one humanoid-maker's story. The KraneShares KOID ETF exists specifically for this layer — a basket built around the actuation, materials, and component supply chain rather than any single robot brand, with Nabtesco, Sanhua, and Lynas all sitting inside it as real, disclosed holdings. It's the practical version of this letter's thesis: exposure to whichever humanoid platform wins, without having to guess which one that will be.
The humanoid robot story is genuinely still speculative at the finished-product level — no dominant platform has emerged, and most of the best-known names (Figure, 1X, Agility) remain private. The component layer underneath it is a different, more mature bet: Harmonic Drive and Nabtesco are established, profitable, decades-old precision manufacturers; Schaeffler's million-unit actuator contract is a real, signed, multi-year commitment, not a projection; and the rare-earth chokepoint this series has flagged elsewhere applies here with the same force, with Lynas as one of the few ways to invest against Chinese concentration rather than into it. This is the pick-and-shovel structure Letter 03 described, transplanted from AI infrastructure to physical robotics — and, per the data, it has been the better-performing half of the humanoid theme so far. The risk to this thesis: component demand is still entirely a bet on humanoid production actually scaling to volume; if the category stalls at demo-stage rather than reaching mass deployment, the suppliers underneath it feel that shortfall just as directly as the robot makers do.
Founder, NextGen Economics · Bangalore, India · August 2026
Sources: Brianartex, "Humanoid Robotics Deep-Dive: Harmonic Drive · Leader Drive · Sanhua · $ROBO" (May 2026) · DataIntelo, "Harmonic Drive System Market Research Report 2034" (Mar 2026) · IntelMarketResearch, "Humanoid Robot Speed Reducers Market Outlook 2026-2032" · KraneShares, "Humanoid Robotics ETF: Top Performing Stocks in the KOID Portfolio" (Mar 2026) · PhotonCap, "Investment Map: 20 Companies in the Humanoid Cycle" — Schaeffler/Humanoid AI deal, Figure AI Helix-02 (May 2026) · Simply Wall St, Harmonic Drive Systems (TSE:6324) investment narrative. This letter evaluates a set of publicly-traded and ETF investment vehicles using real market data; it is not a recommendation to buy or sell any specific security.
A deeper research note comparing component-layer valuations across the full humanoid supply chain, building on the AI infrastructure framework in Letter 03, is available to NGE clients beyond what's covered in this free letter.
Not investment advice. All investments carry risk including loss of capital. Figures cited are sourced from third-party market research and public company disclosures, current as of the dates noted above.