NdFeB Magnets: The Material Bottleneck Behind Humanoid Robotics
TLDR: Humanoid robotics is often framed as a software race, but the binding constraint may be physical. For China, this matters beyond technology: with debt-heavy growth and fewer scalable growth vectors in property and infrastructure, embodied AI is increasingly strategic. High-precision actuators require NdFeB permanent magnets because no other commercially viable material offers comparable energy density at acceptable size and efficiency. The relevant bottleneck for humanoids is not “rare earths” in general, but the NdPr-to-magnet supply chain. China’s dominance in midstream processing and magnet manufacturing gives it leverage over the pace, cost curve, and industrial geography of embodied AI.
Notes:
This article is based entirely on free publicly available information. I do not have access to proprietary sell-side reports, so I’m making a directional case for the thesis.
I strongly recommend reading my previous piece on China’s debt crisis and Taiwan first, since this article builds on it and assumes familiarity with that framework.
The Binding Constraint is Physical
Humanoid robotics is almost exclusively framed as a software race. The consensus view is that once we solve the brain the robots will appear from low cost producers.
This view is wrong.
While the West obsesses over the cognitive bottleneck, the binding constraint for deploying embodied AI at scale is physical. Building millions of humanoids requires a vast ecosystem of materials and capabilities, but the single most critical and constrained variable is torque. High-performance actuators are the only way to achieve fluid movement, and you cannot build them at scale without Neodymium-Iron-Boron (NdFeB) permanent magnets.
NdFeB magnets are the efficiency chip of the physical world. Without them, robots are too heavy to walk and too inefficient to run. And right now, the supply chain that turns raw earth into these high-performance magnets is a veritable monopoly.
China controls >90% of global magnet manufacturing and the complex chemical midstream that makes it possible. For China, this is a macroeconomic escape hatch. Facing a debt crisis and a shrinking workforce, Beijing is betting that embodied AI can be its “Second Lever”, a way to productize labor itself and export it to the world.
China’s Second Lever: The “Hail Mary” for Growth
In my previous piece, I argued that China’s core problem is a debt-heavy system that has exhausted its old growth engines (property and infrastructure). China sits in a growth trap. The debt burden has risen faster than the economy’s capacity to service it, and the country now looks like a system that can’t slow down without breaking something. The deeper binding constraint underneath the headline debt numbers is weak household consumption, which is ultimately a distribution problem. When households capture too small a share of national income, demand stays structurally weak. When demand is weak, growth gets propped up by credit and state balance sheets. And when growth gets propped up by the state, the debt burden migrates into local governments, banks, and policy vehicles. In that sense, the debt problem is downstream.
Escaping the trap without a hard deleveraging requires two things at once:
A productivity shock large enough to stabilize the debt math, keep nominal GDP growing faster than the effective interest burden. Productivity shocks are positive supply shocks: they lower marginal costs, expand potential output, and increase real income (which is what China needs urgently).
A distribution mechanism that routes a meaningful share of that surplus to households, otherwise you just get margin expansion, retained earnings, and a more state-heavy economy with the same demand constraint.
In my previous piece, I argued China needs an economic “hail mary” in the form of a tech-driven productivity shock, something powerful enough to lift output and keep the system solvent without allowing a balance-sheet clearing. To generate that shock, China has two plausible levers:
Lever #1: frontier compute / sub-7nm chips capital-light, high-margin digital growth, but externally constrained by export controls, tooling chokepoints, and the Taiwan nexus.
Lever #2: embodied AI — robots as compensatory capacity and productivity infrastructure that can be scaled inside China’s existing industrial base.
I already covered Lever #1 in the first piece. This article is about Lever #2.
China needs a replacement engine that can (1) lift trend growth, (2) generate tax base and cash flows, and (3) do it without triggering a balance-sheet clearing that produces mass unemployment. The strategic bet is that humanoid robotics, and the upstream actuator stack, can become the next EV/solar-style national champion complex, but with higher strategic leverage because what’s being productized is labor itself.
If China can dominate the humanoid supply chain, especially the actuator corridor where cost, yield, and reliability are set, it can build a globally traded, high value-add export category that throws off profits, payroll, and fiscal revenue. Those flows help keep nominal growth above the interest burden, which is the only non-painful way to shrink a debt overhang in relative terms.
The Trade War Shifts From the Brain to the Body
China’s problem is not necessarily an innovation in the abstract. It’s finding a growth vector that is (1) scalable inside China’s factor endowments, (2) hard to sanction, and (3) capable of throwing off durable nominal cash flows to outrun the interest burden. Frontier compute is the best answer on paper (as argued in my first article), but the problem is that it’s externally gated by export controls, tooling chokepoints, and the Taiwan nexus.
But embodied AI is the opposite: capital-deepening, industrial, and physically buildable at home.
That’s why Beijing treats humanoids like productivity infrastructure, a way to convert China’s enduring strengths (industrial clustering, supply-chain density, scale manufacturing, state-directed capital, and process learning) into a new export engine.
The West mostly treats humanoids as a software problem: solve the model, and the body becomes a commodity. Beijing treats humanoids as an industrial strategy: secure the body, and you control the learning curve, the cost curve, and the exportable unit of labor.
It absorbs balance-sheet capacity without the old pathologies. Property and infrastructure were capital sinks that increasingly produced low social returns. A robotics buildout is still investment-heavy, but it’s tethered to tradable output, process learning, and export revenue.
It addresses the demographic constraint in the least politically painful way. A shrinking workforce and rising dependency ratio are a growth-killer. Robots are synthetic labor supply, and a way to raise output per worker and defend unit labor costs in the sectors that feed directly into living standards.
It creates an externally monetizable surplus. Weak consumption means China can’t rely on its domestic demand to clear ever-higher production. Humanoids can become a globally traded capital good that can be exported into other countries’ labor shortages. That’s exactly the kind of high-margin growth vector you want when the home market is demand-constrained.
My core claim is that the competition is shifting from the brain (semiconductors, frontier models, training infrastructure) to the body (actuation). If the U.S. does not own that corridor, it will import the bottleneck module that gates volume. And if you import the bottleneck, your learning curve compounds inside the exporter’s cluster, not domestically, which is exactly how China locked in advantage in solar, batteries, and increasingly EV subsystems.
In that framing, NdFeB (core input in actuators) is the strategic material that sets the ceiling on embodied AI throughput. That’s why the “body” matters. The next decade will be decided by who can manufacture the platform the model inhabits.
If the macro strategy provides the motivation, the machine itself provides the means. But why the humanoid form factor specifically? Why not simply deploy more traditional industrial arms?
Humanoids: Why They Matter
Humanoid robots are a bet that general-purpose labor can be productized. The reason the humanoid form factor matters is because the world is already built for humans. Doors, stairs, shelves, tools, carts, and factory layouts all assume a human body plan. If a machine can operate inside that environment without requiring the environment to be redesigned, it becomes deployable across many jobs.
That is what makes humanoids economically different from traditional industrial robots. Most industrial robots are task-specific and require a controlled setup. Humanoids aim to reduce that integration burden.
The timing is also important. For decades, most advanced economies benefited from a demographic dividend, more working-age people meant more output, more consumption, and a stable tax base.
That era is ending. Across the OECD, the old-age dependency ratio (65+ per 100 working-age people) has already risen from ~19% (1980) to ~31% (2023) and is projected to reach ~55% by 2060 across the OECD. In simple terms, fewer workers must support more retirees, just as demand for healthcare, logistics, and elder care rises. In that context, humanoids are the best pathway to convert AI from digital productivity into physical capacity and bridge that shortage.
But to build them at the scale a plethora of materials and components are needed.
Components
After that framing, it helps to ground the reader in what a humanoid actually is. This is, of course, a very simplified map of an extremely complex machine (if you want to dive deeper, take a look at Humanity’s Last Machine).
At a high level you can think of five buckets:
Structure is the skeleton (frames, housings, castings) that carries load and survives impacts.
Energy is metabolism (the battery pack and power electronics) that sets runtime and peak power.
Sensing is eyes and ears (cameras and other vision sensors) that let the robot perceive the world.
Signaling is the nervous system (wiring, connectors, and force/torque sensing) that moves information and feedback through the machine.
Motion is the equivalent of joints and muscles, where electrical power becomes controlled movement at every joint.
It’s easy to over-index on the software bottlenecks like quality data, world models, and the long tail of edge cases. But they are ultimately cognitive bottlenecks, they improve with iteration, better data engines (simulation + teleop + self-supervision), and scaling compute.
Physical constraints behave differently. Hardware scaling is gated by throughput, yield, heat, wear, and supply chains, constraints that are slow to relax because they require new factories, new processes, and years of ramp. That’s why the most important question for humanoids is if the body can be built reliably, millions of times, at an acceptable cost?”
In practice, the answer is determined by the robot’s motion hardware at the joints: it’s the highest-cost, steepest-learning-curve subsystem, and the one that most directly sets performance, manufacturability, and scalability. That joint-level motion package is the actuator.
What Is An Actuator ?
Think of an actuator as the robot’s muscle at each joint. As Humanity’s Last Machine puts it, “a humanoid robot can be thought of as a collection of actuators operating in coordination”.
When a humanoid bends an elbow, lifts a leg, or rotates its torso, something has to (1) create force, (2) control it precisely, and (3) survive repeated stress. That whole “muscle package” is the actuator.
Rotary vs linear
In humanoids, you’ll typically see two ways of making a joint move: rotary actuators and linear actuators. The diagram below shows this. R joints (rotary) are everywhere you need something to spin like a hinge (think shoulder rotation, elbow bending, wrist turning). L joints (linear) show up where it can be easier to generate force by pushing/pulling in a straight line, and then converting that straight motion into joint rotation through linkages.
The important message from the diagram is that a humanoid is composed of dozens of small, specialized force-generators distributed across the body, some delivering rotation (R) and some delivering push/pull (L), but all serving the same goal of generating controlled force at the joints.
Electric vs hydraulic vs pneumatic
Actuators can be powered electrically, hydraulically, or pneumatically. Hydraulics offer very high power density and are excellent for brute-force motion; pneumatics can be cheap and clean but struggle with precision and controllability. But for general-purpose humanoids the industry trend is overwhelmingly electric actuation, because it’s compact, programmable, repeatable, and easier to integrate with batteries and onboard electronics.
What’s inside an electric actuator
An electric actuator is essentially a compact motion package built from a few parts:
Motor: generates torque (this is where permanent magnets matter)
Gearbox / reducer: trades speed for torque (and improves controllability)
Bearings: carry loads and keep the joint stable
Encoder / sensors: tell the controller where the joint is and how it’s moving
Housing: keeps everything aligned and rigid under stress
Of these, the motor is the most crucial and typically the most expensive single element to get right, because it sets the ceiling on torque density, efficiency, heat, and size. And that’s exactly where NdFeB permanent magnets enter. They’re what enable high torque in a small package at acceptable efficiency. If the magnet is weaker, you can still build a motor, but you pay for it in size, weight, heat, and battery drain, which compounds across dozens of joints.
Actuation is the cost base
The Optimus-style breakdown makes the point visually. A humanoid is effectively a battery strapped to a distributed network of actuators. The actuator-heavy limb and joint modules add up to ~$52.3k, or ~95% of the total. The exact numbers will vary by design, but the message is clearly that humanoids scale as a repeated stack of motors + reducers/screws + bearings + sensors across the body (with NdFeB magnets alone representing ~15–20% of total build cost).
And because the motor is the most complex and expensive part of each module, and the motor’s performance depends disproportionately on the magnet, the humanoid scaling question quickly collapses into a materials question. That’s why we now zoom in on NdFeB.
Where NdFeB fits
At the heart of every actuator lies a NdFeB permanent magnet as discussed previously. These magnets represent a fundamental step-change in energy density, achieving roughly 400 kJ/m³, the highest of any commercially viable magnetic material.
Rare-earth magnets sit in a different performance regime than conventional materials, enabling high torque in compact packages, exactly what humanoid joints and precision actuators require. If a humanoid robot were built with conventional magnets, its motors would need to be 4x larger to generate the same torque. The robot would become so heavy it could barely lift its own limbs, let alone a payload.
The Value-Volume asymmetry
This also explains why the rare earths debate is usually framed too broadly. We must look at the disparity between volume and economic value. Rare earths are used in a wide range of industrial applications, from catalysts and glass polishing to ceramics, but their strategic importance is not distributed equally.
By volume, permanent magnets represent roughly 35% of total rare earth consumption. However, when viewed through the lens of economic value, they account for a staggering 91% of the market. This massive concentration of value reflects the fact that high-performance magnets (NdFeB) require the highest-purity inputs, specifically Neodymium, Praseodymium, and Dysprosium, which command the highest prices and face the greatest supply constraints.
Scaling Math
All of that explains why magnets matter.
The next step is why they become the binding constraint under scale. The cleanest way to see it is a stress test: if humanoids ever reach mass deployment, which input becomes the outlier? The chart below ranks bottlenecks by showing how many multiples of today’s annual production would be required under an optimistic deployment scenario.
In this scenario, NdFeB magnet output would need to expand by roughly 186×, versus 14× for lithium, 13× for graphite, and single-digit multiples for cobalt, nickel, and copper.
This is the definition of a binding constraint. While the battery supply chain needs to scale by an order of magnitude, the magnet supply chain needs to scale by two orders of magnitude. NdFeB is the outlier input, and therefore, the pace of humanoid scaling will be gated by the pace of magnet production.
Why Standard TAM Models Fail
Most analysts try to size this market by multiplying “Volume x Price.” They assume that as robot production scales, component costs will fall (Wright’s Law).
For batteries and sensors, that is true. For magnets, it is likely false.
If NdFeB supply is structurally inelastic (as I will explain in the next section), then a rapid humanoid ramp creates a shortage. In a shortage, the price of a bottleneck component does not track its cost of production; it tracks its marginal value to the buyer.
The Baseline: Currently, NdFeB magnets represent ~15–20% of a humanoid’s Bill of Materials (BOM).
The Squeeze: If 186x demand hits a fixed supply base, the price of NdFeB will decouple from the cost of mining. It will rise to the price of the next best alternative (which, in robotics, is a heavier, less efficient robot that can’t do the job).
Therefore, standard TAM models that predict a “linear” growth in magnet revenues are likely underestimating the violent repricing that occurs when an exponential demand curve hits a vertical supply curve.
TAM Estimate
To translate the humanoid revenue forecasts into a “magnet TAM” number, you need one simplifying assumption: NdFeB magnets represent ~15–20% of the total build cost of a humanoid (as discussed above). If that’s directionally right, then a rough first-pass TAM is just:
NdFeB-linked TAM ≈ Humanoid market revenue × (15–20%)
Using the two exhibits below (US humanoid revenue in USD; China humanoid market size in RMB) and converting China at roughly ~7.2 RMB/USD, we get an US + China proxy:
2030: US ~$4B + China ~RMB 12B (~$1.7B) ≈ $5.7B
NdFeB-linked TAM ~$0.9–$1.1B
2040: US ~$238B + China ~RMB 1,043B (~$145B) ≈ $383B
NdFeB-linked TAM ~$57–$77B
2050: US ~$1,001B + China ~RMB 6,092B (~$846B) ≈ $1,847B
NdFeB-linked TAM ~$277–$369B
A useful comparison is Morgan Stanley’s “Estimated Annual Demand – Magnets” table below, which pairs a large volume ramp with a relatively tame pricing path. In their model, the average cost of rare-earth magnets rises from ~$50/kg in 2024 to ~$100/kg by 2050, a clean 2× over two and a half decades, while the total rare-earth magnet market scales from ~$142M (2024) to ~$1.6B (2030), ~$10.8B (2035), ~$44.7B (2040), and ~$166.8B (2050). My view is that this likely understates the price lever. If NdFeB supply is structurally inelastic (as I will explain later) then a rapid humanoid ramp doesn’t just increase tonnes; it forces the market to clear via sharper price moves during tightness episodes.
So, even if overall humanoid manufacturing costs compress over time (learning curves, scale, cheaper electronics and other components), NdFeB can still become a larger cost driver of the BOM because it’s one of the few critical inputs whose supply response is slow. That’s also why I suspect my earlier ~15–20% NdFeB-related cost share may be conservative: if magnet $/kg rises faster than the rest of the robot cost stack falls, NdFeB’s share of cost can creep upward as production scales.
Before talking about supply and inelasticity, we need to pin down the demand side in units. The next section quantifies NdFeB demand bottom-up to show why the coming constraint is likely tighter than what sell-side analysts imply.
The Demand Landscape: A Zero-Sum Game
As mentioned previously, NdFeB is the only commercially scalable option for humanoids. That immediately makes the demand landscape a zero-sum game: every incremental kilogram of NdFeB pulled into robotics is a kilogram that can’t go into EV drivetrains, wind generators, or industrial motors unless the supply chain expands—which it doesn’t, quickly.
Most published forecasts point in the same direction: NdPr and NdFeB markets are structurally tight. The disagreement is not whether there is a deficit—it’s what drives it and how large it becomes once robotics moves from “niche” to “deployment.” By 2040, several models already imply an annual NdFeB undersupply on the order of hundreds of thousands of tonnes even after assuming mine restarts, capacity expansions, and recycling gains.
But the bigger issue isn’t the headline demand number—it’s the demand mix. And that’s where most forecasts fail.
Demand Mix: The Missing Category
Most demand models implicitly assume robotics remains marginal. You can see this in their category design: robotics is absent as a standalone line item and gets folded into broad buckets like “industrial motors” or “other sintered magnets.” That choice isn’t neutral—it hard-codes a conclusion. If robotics is not modeled explicitly, it is assumed to be immaterial and therefore incapable of reshaping the demand curve.
In this forecast, the world reaches demand of ~540 kt by 2038 without ever allocating a line item to robotics, meaning that robotics is assumed to be either de minimis or embedded inside electronics/other. The graph correctly forecasts a steep ramp driven by NEVs and wind turbines though.
Even in the most aggressive forecast I found, robotics remains capped at <20% by 2050, despite robotics being the only category that scales multiplicatively through actuator count. This is, in my view, a complete underestimation.
The Policy Blindspot
The scariest part of this math is that Western policymakers aren’t even tracking it.
Two of the most authoritative frameworks, the Section 232 investigation report (The Effect of Imports of NdFeB Permanent Magnets on National Security) and the U.S. The Department of Energy’s 2022 supply-chain deep dive (Rare Earth Permanent Magnets), don’t even include robotics as a distinct demand category. Their demand tables allocate the future mostly across wind and EVs, while everything else gets buried in broad buckets like “industrial motors” or “other sintered magnets.”
When robotics is absent from the baseline, policymakers are implicitly planning for an EV/wind-only world, under-building the magnet corridor relative to the actual embodied-AI demand path.
The disagreement is therefore about how robotics scales. To bound the problem, we can reduce robotics-driven magnet demand to three variables. If any one of them is understated, the forecast misses. If all three are understated the miss compounds.
From here, the entire debate collapses into three variables:
Number of robots (humanoids, service, industrial, etc…)
magnet mass per actuator (torque density and thermal performance constraints)
actuators per robot (degrees of freedom, dexterity, and task coverage).
Number of Robots
Base case (2050 = 100M units/year): this anchor is deliberately back-of-the-envelope and comes from extrapolating the line from Android Dreams that “by 2040 Chinese humanoid companies are churning out humanoids by the millions per week.” “Millions per week” implies an annualized run-rate on the order of ~50M–200M+ per year (e.g., 1–4M/week ≈ 52–208M/year). We take 100M/year as a midpoint-ish, still conservative base case within that implied band, especially because the quote is specifically about China’s production capacity by 2040, and 2050 would plausibly include global scale-out.
Scenario bounds (2050 = 50M / 100M / 200M): the conservative and upside cases bracket the same interpretation—50M/year is “one-million-per-week-ish” in steady state, while 200M/year reflects a world where the “millions per week” statement becomes a durable industrial reality and scales beyond a few champions.
China’s adoption curve (shown below) is a useful reality check: it ramps annual new units from near-zero in the 2020s to ~59 million humanoids per year in China alone by 2050. Against that anchor, assuming the rest of the world adds only ~40 million incremental units per year to reach a ~100 million/year global base case is very conservative.
Actuators / Humanoid
The main driver of rising magnet demand per robot is the push toward higher degrees of freedom (DoF). Today’s humanoids are still mechanically sparse relative to a human: while many current platforms sit in the ~30–50 joint range, the human body has ~78 joints. For the 2030 conservative anchor (28 actuators), we’re using Tesla Optimus’ 28 actuators. That makes 28 a plausible “downside” assumption for 2030: early mass-deployment robots will likely optimize for cost, reliability, and manufacturability rather than full human-like articulation.
The hands are the clearest indicator of where the ceiling goes. Dexterous manipulation is actuator-hungry, and some robotic hand designs already reach ~22 DoF per hand, while humans have 27, implying that humanoids will likely add actuators disproportionately in the upper body even before full-body joint parity is reached. Over the long run, the upside case is humans exceeded: robots can justify additional joints/actuators that improve task coverage and controllability. For example, more wrist axes, extra finger articulation, or redundant DoF that makes grasping robust across object shapes and failure modes. Beyond hands, designers can also add or expand articulation in areas where humans are constrained by anatomy (or where robot form factors diverge): more freely rotating heads/neck axes for sensor pointing, added hip/waist axes for load handling and reach, and other non-biological joints that trade marginal actuator cost for gains in mobility, stability, and workspace.
NdFeB Intensity / Humanoid
We should expect downward pressure on NdFeB per actuator as prices rise, engineers will thrift magnet mass wherever they can. But the Cost of Transport (CoT) constraint means it can’t fall very far in a humanoid. Every gram added at the extremity cascades up the kinematic chain: a heavier ankle actuator forces upsizing at the knee, then hip, then battery, and a ~200g “error” can compound into a ~1.3kg system-level penalty. That creates relentless incentives to minimize magnet mass per actuator even as actuator counts climb toward (and potentially beyond) human-level dexterity.
At the same time, designers can’t simply substitute away from NdFeB: weaker magnets require higher current to hit the same torque, driving resistive losses and heat, wasting battery, and increasing demagnetization risk at elevated temperatures. Net: more actuators per humanoid, somewhat less NdFeB per actuator, so kg NdFeB per humanoid rises over time—but sublinearly, not a clean 2× step-up just because actuator count increases.
Bottom Line
Anchoring the total NdFeB magnet market (kt/yr) to the Thunder Said forecast, a simple humanoid-only thought experiment already implies a regime change in demand mix. In our base case, humanoids rise from effectively immaterial in 2030 to ~37% of total NdFeB demand by 2050 (500 kt out of ~1,350 kt). In a blue-sky trajectory—where both annual humanoid shipments and per-humanoid NdFeB intensity compound—humanoids alone can plausibly approach ~90% of total demand by 2050 (1,200 kt out of ~1,350 kt).
The uncomfortable part is that major “authoritative” policy-style demand frameworks (as shown above) don’t even include robotics/humanoids as a standalone category, which implicitly assumes it stays small—and therefore can’t reshape the curve. But even if you think our assumptions are aggressive, the directional point survives: humanoids alone can become a first-order driver of NdFeB.
And bear in mind this is only humanoids. If you layer in other robotics categories (industrial robots, AMRs/AGVs, professional service robots, etc.), total robotics-related NdFeB could reasonably be another ~2–3× the humanoid-only number depending on unit volumes and motorization intensity—so this “humanoid” math should be read as a lower-bound stress test, not a full robotics demand model.
Demand is only half the equation. Once you accept that robotics can become a first-order driver of NdFeB, the next question is mechanical: can the supply chain expand fast enough to meet it? And this is where the story turns from “big TAM” to “binding constraint.” The NdPr-to-magnet corridor is not a commodity pipe you can widen with price alone—it is a slow, regulated, chemically intensive industrial stack.
Supply Chain
The NdFeB magnet supply chain is a multi-tiered industrial ecosystem that transforms mineral-rich ores into high-performance permanent magnets essential for the modern energy transition. The system is fundamentally divided into upstream extraction and separation, midstream metallization and alloying, and downstream magnet manufacturing and finishing.16 NdFeB magnets are comprised approximately of 30% rare earth elements (primarily Nd and Pr, with Dy and Tb used for high-temperature grades), 69% iron, and 1% boron.
Mining (Ore Extraction)
Rare-earth ores (like bastnäsite, monazite or ionic clays) containing Nd and other REEs are mined mostly in China. In 2024 China accounted for ~69% of global rare-earth oxide (REO) production, dwarfing all other countries. But mining is the least unequal part of the chain: the U.S. and Australia do produce meaningful ore, and the real asymmetry shows up later, during separation, metal/alloy, and magnet manufacturing.
As seen, other leading producers include the United States and Myanmar (Burma) at far smaller shares. China’s massive Bayan Obo iron-rare-earth mine (Inner Mongolia, by China Northern Rare Earth Group) is the world’s single largest source of REO. Outside China, major operations are fewer. The U.S. has one large rare-earth mine at Mountain Pass, California (operated by MP Materials), a bastnäsite deposit that produced roughly 45,000 t/year of REO in 2024, which is enough to produce ~ 13% of current world Nd/Pr oxide. Australia’s main rare-earth producer is Lynas Rare Earths, which operates the Mt Weld carbonatite deposit and produced ~10,000 tonnes in 2025, about 63% of which was NdPr oxide.
To make it concrete, compare the images below: Bayan Obo (China) versus Mountain Pass (U.S.), which are the world’s largest and second-largest rare earth mines respectively.
Therefore, mining is the one stage where the West still shows up, Mountain Pass and Mt Weld prove there’s sufficient ore outside China. But the choke point starts right after mining, in the separation stage, where mixed concentrate gets split into high-purity NdPr (and Dy/Tb) oxides. That’s where China’s control becomes overwhelming.
Separation (Chemical Processing)
After mining, rare-earth ore concentrates are chemically processed to separate individual rare-earth elements (or groups) from each other. In practice this uses solvent-extraction: the ore concentrate is leached (often with acid), then passed through stages of solvent exchange where lighter REEs (La, Ce, Nd, Pr) are separated from heavier ones (Dy, Tb, etc). This is a complex, multistep chemical process (hundreds of mixer-settler operations) that requires large amounts of reagents and water. Uranium and thorium impurities in the ore must also be removed, making the chemistry more challenging.
China controls ~90% of global rare-earth separation capacity, with large state and private separation plants, linked to major mining districts like Bayan Obo handling most of the world’s concentrate. Malaysia is the most important non-China separation node today via Lynas Rare Earths’s LAMP facility in Kuantan, which processes Mt Weld concentrate and has expanded capabilities over time. India has IREL Limited, and Estonia represents the newer “downstream comeback” in Europe. Neo Performance Materials opened a rare-earth permanent magnet manufacturing facility in Narva in 2025 to reduce reliance on Chinese magnet imports, supported by EU funding. Beyond those four anchors, the rest of the non-China ecosystem is still relatively thin and fragmented.
Refining (Oxide-to-Metal Conversion)
Separation yields rare-earth oxides (REO). The refining stage converts these oxides into either pure RE metals or alloy powders for magnets. There are two main routes: reduction (chemical or electrochemical) of oxides to metal, and alloy melting. For NdFeB magnets, Nd and Pr oxides are typically reduced with calcium (or aluminum, etc.) in high-temperature furnaces to make Nd–Pr alloy metals, which are then mixed with iron and boron and melted into NdFeB alloy ingots. (Dy and Tb are often added as ferroalloys if needed for high-temperature performance.)
Nearly all RE metal refining is still done in Asia. China produces again ~ 90% of rare-earth metal and alloy output. A few non-Chinese facilities exist: Vietnam Rare Earth JSC (Vietnam) runs a plant to make NdFeB alloy, and Less Common Metals (UK) and Silmet (Estonia, now part of Canada’s Neo Performance Materials) have limited separation-to-metal lines. Newer efforts to rebuild refining include: U.S. company MP Materials plans to add oxide-to-metal refining at Mountain Pass, and Energy Fuels is studying full refining of its Utah monazite output. South Korea’s LS Cable/LS Eco Energy is investing in a rare-earth metal refinery in Vietnam to integrate imported oxides into NdFeB metal.
Magnet Manufacturing
Finally, refined Nd–Pr–Dy (and other metals) are alloyed with iron and boron to make NdFeB magnets. The alloy is cast and machined into ingots or pellets, then either sintered (pulverized into powder, pressed in a magnetic field, and heat-treated) or bonded/injection-molded into final magnet shapes.
Again, China overwhelmingly dominates, they produce ~92% of the global market. Outside China, Japan is the largest producer (~7% market share), with several long-established firms: Hitachi Metals, Shin-Etsu Chemical, and TDK, which make magnet alloys, powders, and finished magnets. Other notable producers include Vacuumschmelze (Germany/Apollo-owned) and its Finnish arm Neorem, which supply high-performance magnets to Europe, and Neo Performance Materials (Canada), which makes bonded magnet alloys. The UK’s Less Common Metals supplies NdFeB powders (mostly for bonded magnets).
In the United States, domestic NdFeB magnet production has been negligible; until recently the only U.S. plant was Urban Mining Company (Texas), which recycles scrap NdFeB into sintered magnets. New U.S. projects are changing this: in 2021 MP Materials announced a magnet factory in Fort Worth, TX (with General Motors) to make ~1,000 tonnes/year of NdFeB magnets from its Mountain Pass output. In 2023, Vacuumschmelze also agreed to produce magnets in the U.S. for GM.
Even if the U.S., Europe, and Asia ex-China all execute on the announced buildout (which I highly doubt), the scale gap doesn’t really close. By 2030, the “rest of world” is only ~50–55 kt/year versus ~280 kt/year in China, so China still sits at roughly >82% of global NdFeB magnet capacity by 2030.
And it’s not surprising this is where the West gains the most traction, magnet manufacturing is the easiest step to onshore relative to mining, refining and separation. It’s more modular, capex is lower, permitting is less politically explosive, and you can build plants near end-demand (EVs, wind, robotics) without needing to solve the nastiest chemistry upstream. So even if you win share in this box, you still need the NdPr (and Dy/Tb) inputs, and those bottlenecks sit upstream, where China is and will remain dominant.
The Ends Diversify, the Middle Stays Chinese
At a high level, the NdFeB chain is bifurcated: China can lose share at the ends, but it still dominates the middle.
My base case is that China loses share at the “ends” of the chain but maintains (or even gains) share in the two middle links. Mining is the most globally contestable step because ore bodies exist outside China and incremental supply can be brought online without immediately recreating the downstream chemical stack; that pushes China’s mining share down over time. Magnet production is also relatively onshorable: plants are modular, capex is manageable, and governments can subsidize capacity close to demand in EVs, wind, and defense, so the U.S., Europe, and Asian allies can take some share here.
But the hard constraint is the middle. Separation and refining are the chemistry-heavy, scale-driven steps with the highest compliance burden and the steepest learning curve, and they remain dominated by China. As a result, even if the West wins share in magnet manufacturing, it still needs NdPr (and Dy/Tb) oxides/metals/alloys, which means the dependency often just migrates upstream. In practice, that dynamic can actually increase China’s leverage in separation/refining even as its share slips in magnets, because whoever controls the chemical conversion controls throughput, pricing, and who gets material when markets tighten.
If the chain were elastic, a price spike would trigger a visible ramp in NdPr/Dy/Tb output within a few quarters. Instead, history shows the opposite, when prices surge throughput barely budges. Three episodes make that clear.
Analysis of Supply Inelasticity via Historical Episodes
The 2010–2011 Senkaku Price Shock
The first major rare earth crisis was precipitated by a territorial dispute between China and Japan, leading Beijing to weaponize its market dominance. Between January 2009 and August 2011, the price of NdPr oxide experienced a ~5-fold increase, soaring from $38/kg to $190/kg.
During this 31-month rally, non-Chinese supply was almost completely unresponsive. In 2010, China controlled 97% of all rare earth mineral production. The extreme price signals failed to bring new mines online because the necessary infrastructure for separation and refining outside of China had been dismantled in the previous two decades due to low-cost competition. Instead of a supply-side ramp, the market experienced “engineered demand destruction,” where manufacturers like Tesla and Toyota began researching magnet-free motors or induction systems to mitigate supply risk.
The 2021–2022 Post-COVID Crunch and Energy Crisis
The second episode occurred as the global economy emerged from the COVID-19 pandemic. NdPr oxide prices spiked from approximately $45/kg in 2020 to a peak of $170/kg in early 2022. This 375% increase was driven by a combination of surging EV demand and supply-side disruptions, including border closures with Myanmar and power cuts in Southern China.
Quantitatively, the supply response remained muted. While China raised its official mining quotas by 25% to manage the market, non-Chinese producers such as MP Materials in the U.S. were already operating at maximum capacity. Because the lead time for new mine development is 10 to 20 years, no significant new entrants could capitalize on the high prices. The inelasticity was further underscored by the fact that even with record-high prices, global rare earth mine production increased by only a fraction of the demand growth, leading to a persistent supply-demand gap.
The (2023 - Present) Strategic Export Control Regime
This episode is policy-driven scarcity hitting a capacity-constrained midstream. Non-China separation/refining/alloying is still thin and only beginning to ramp, so any tightening upstream quickly translates into shortages downstream.
In 2025, China added export licensing for rare-earth magnets and processing equipment. NdPr prices were up ~30.9% YoY by late Oct 2025. Western supply remained almost fixed due to a lack of investment and fiscal scaffolding (until recently). Without durable tools (offtakes, price floors, capex support), projects struggle to clear financing through cycles. That fragility showed up in late 2023 when environmental licensing delays pushed out roughly ~10,000 tons of planned non-China expansion, roughly ~20% of the global NdPr market, effectively neutralizing the supply response even as prices moved.
To summarize:
Taken together, these three episodes show the same pattern: when the system is stressed, NdFeB supply doesn’t respond like a normal commodity. Prices can move violently, but non-Chinese output ramps slowly, if at all, because the bottlenecks sit in processing, equipment, permitting, and manufacturing learning curves, not in the orebody itself.
So the forward-looking question becomes: what exactly makes this supply chain so hard to scale, even when economics say it should? The answer is not one constraint but a stack of them.
Main Sources of Inelasticity:
The persistent failure of supply to meet demand is rooted in twelve specific structural bottlenecks across the upstream, midstream, and downstream phases of the value chain.
Upstream Phase: Extraction and Separation
1. Separation Is Chemical Sorting at Industrial Scale:
Rare earths are difficult to process because the elements are chemically very similar, so you can’t separate them in a few straightforward steps like copper or iron. Instead, producers rely on a long chemical sorting line called solvent extraction, which can involve 1,000+ linked mixer-settlers working in sequence. Reaching high purity (around 99.5%+) isn’t instant either, the system often needs days to weeks to fully stabilize.
2. Environmental Permitting and Radioactive Waste Management:
Neodymium’s processing emissions are shown at roughly ~70–80 tCO₂ per tonne on the chart, which is about ~8× higher than most other major metals on the same axis. Rare earth ores such as monazite and bastnäsite are often co-located with thorium and uranium, so scaling NdPr is not only a mining problem but a regulated waste-handling problem: radioactive solids plus large volumes of acidic/contaminated wastewater. That combination translates into heavier environmental review, more community pushback, and fragile timelines. So even if feedstock exists, Western projects can stall at the tailings and disposal bottleneck.
3. Protracted Project Lead Times:
The timeline for critical mineral projects is fundamentally incompatible with the fast-moving demand cycles of the EV and electronics industries. Lead times from initial geological discovery to commercial oxide production range from 15 to 20 years.7 The exploration, pilot testing, environmental impact assessment, and construction phases are sequential and cannot be compressed by capital alone. Consequently, supply in 2025 is largely determined by investment decisions made in 2010, and investments now will dictate supply in 2040.
4. The By-product and Ratio Mismatch Problem:
Neodymium and praseodymium are rarely found in high concentrations without a surplus of less valuable light rare earths like cerium (Ce) and lanthanum (La). Neodymium typically accounts for only 15-20% of the total rare earth content in an ore body. When demand for magnets drives a need for more NdPr, mines must also overproduce Ce and La, which currently suffer from near-zero margins. This creates an economic drag where the revenue from NdPr must subsidize the loss-making extraction of unwanted by-products, effectively raising the price floor required for new mines to be viable.
Midstream Phase: Refining and Alloying
5. Geographic Concentration of Refining and Metallization:
While mining has diversified slightly, the “midstream” of refining and metallization remains a virtual monopoly. China manages approximately 92% of global refined NdPr oxide. This concentration creates a strategic bottleneck where non-Chinese miners are forced to ship their concentrate back to China for processing because domestic refining capacity is absent or undersized. This circular dependency negates the elasticity of Western mining efforts.
6. Specialized Equipment and Component Scarcity:
The manufacturing of magnet-grade alloys requires specialized equipment, including vacuum induction furnaces, hydrogen decrepitation (HD) chambers, and jet mills. As of 2025, China has escalated controls on the export of this equipment, specifically targeting technologies used to produce permanent magnets. This prevents new market entrants from procuring the high-efficiency production lines needed to compete on yield and cost, further insulating the supply from price-driven expansion.
7. Capital Cost Asymmetry and WACC:
Rare earth projects are extremely capital-intensive, with a single low-capacity separation plant requiring $500 million to $1 billion in capex. In the West, the WACC for these projects can drift into the ~12–15% range, while China offers ~2% China financing for strategic sectors (I.e. rare earths).
This is because the West’s monetary framework is built around targeting CPI and not the cost of capacity (the cost of building the mines, separation plants, smelters, and factories). This leads to the Fed raising rates when inflation rises. But critical-mineral projects are long-duration bets, so higher rates hit them twice, by raising interest expense on multi-year construction debt and by pushing discount rates up on long-dated cash flows. Therfore, tightening policy to fight today’s inflation can destroy the supply-side investment that would prevent tomorrow’s shortages.
8. Scarcity of Seasoned Technical Human Capital:
The metallurgy of rare earth magnets is a niche field with a severely limited global workforce. Outside of China and Japan, there is a chronic shortage of process metallurgists and quality assurance engineers specialized in NdFeB production. So even if the west wants to build capacity and establishes the proper fiscal framework around it, there is a significant talent bottleneck.
Downstream Phase: Magnet Manufacturing and Finishing
9. Manufacturing Yield:
NdFeB production involves a powder metallurgy route that is highly sensitive to oxygen and temperature. Early-stage facilities typically achieve only 70-80% first-pass yields, whereas mature Chinese facilities operate at 90%. Scrap material becomes an expensive waste stream, and the reprocessing of magnet scrap adds a 15-25% cost premium. These yield challenges mean that even when a factory reaches nameplate capacity, its effective output of sellable magnets is significantly lower for the first several years.
10. Mechanical Brittleness and Handling Failures:
NdFeB magnets are inherently brittle and exhibit significant rate sensitivity in their compressive behavior. They are prone to spontaneous collapse and fracture under mechanical stress or impact. This brittleness imposes significant constraints on the assembly and machining phases, leading to high failure rates in automated manufacturing lines. The technical difficulty of handling these “brittle ceramics” prevents the rapid scaling of high-speed assembly processes, acting as a final drag on supply-side elasticity.
Who Captures Value Under Scarcity
If NdFeB were just another commodity, the winners would be the miners. But in a shortage regime, value flows to the narrowest part of the chain.
The profit pool migrates toward the layers with (i) the most inelastic supply, (ii) the highest barriers to entry, and (iii) the greatest ability to pass cost shocks downstream. That is the midstream: separation, refining, and magnet manufacturing .
China controls roughly ~91% of separation/refining and ~94% of NdFeB magnet production. A clean way to see this pricing power is to watch how feedstock shocks propagate. When NdPr oxide prices jump, the midstream behaves like a pass-through buffer, instantly repricing alloys to protect margins while downstream OEMs are forced to absorb the cost
The Gravity Well: Clustering Logic
If NdFeB were abundant and globally fungible, robot manufacturing would locate where labor is cheapest, demand is closest, or incentives are highest. Under structural shortage, that logic flips. When a single input becomes quantity-constraining and clears through allocation, the dominant objective is supply assurance. The predictable result is industrial clustering, where the downstream manufacturing migrates toward the geography that controls the constrained input (NdFeB magnets).
Today, that geography is China. China’s dominance in the midstream means it can deliver magnet-grade material at scale, with shorter lead times and higher reliability than anywhere else. And because humanoids are actuator-dense and complex to manufacture, any disruption in magnets is a production stop. When production-stop risk is existential, OEMs rationally bias toward the ecosystem that maximizes continuity of supply, engineering support, and ramp execution. That is a major reason the robot industrial stack increasingly co-locates in China.
Once supply assurance becomes the key variable, the market naturally moves toward vertical integration. I expect upstream firms to keep moving downstream, either directly, or through JVs and M&A, because embedding a scarce input inside a harder-to-substitute component increases pricing power and deepens customer dependence. So, if the magnet supplier also supplies the rotor, and the rotor supplier co-develops the actuator, the robot OEM becomes dependent on an integrated cluster that can guarantee volume, quality, and iteration at scale.
Thus, if the U.S. wants to reshore robot manufacturing it can’t just “incentivize assembly” and hope the rest follows. The only way to unwind that gravitational pull is to recreate the constrained NdFeB supply chain domestically: secure magnet-grade NdPr throughput, build qualified NdFeB magnet capacity, and scale actuator manufacturing. Once that exists, the rest of the stack can follow.
This clustering assumption is what lets us zoom out from “magnets” to “humanoids” more broadly. Once a single input gates throughput, the bottleneck becomes the control point for the entire learning curve. In humanoids, that control point is the NdFeB → motor → actuator corridor. So when I talk about “who wins humanoids,” I’m not making a generic claim about AI talent or software, it’s a narrower industrial claim: the geography that controls actuator throughput will control the pace of embodied AI scaling, because it owns the module that gates volume, reliability, and cost.
So what does “break the cluster” actually mean mechanically? It means owning the module where costs and reliability compound fastest
Actuators are the flywheel
Actuators are the humanoid’s scaling unit: they’re the most complex, most expensive subsystem, and the one that ultimately determines whether robots can move reliably at scale. Inside that stack, NdFeB magnets are the most supply-constrained, high-value input. Put differently: if humanoids become a mass market, the industrial gravity well forms around actuators, and around the NdFeB corridor that feeds them.
The key mechanism is manufacturing throughput. Once the software stack is good enough, the compounding advantage shifts to whoever can manufacture, qualify, and iterate the physical joint modules fastest. That’s why catch up everywhere is not a plan for the U.S. The only viable response is focus. The U.S. should pour disproportionate resources into the one vertical that governs the learning curve, the cost curve, and the reliability curve for humanoids, then force a manufacturing growth loop there through automation, subsidies, and scale.
Robotics has an exponential, self-reinforcing manufacturing curve, just as AI research can create a self-reinforcing intelligence curve. Once a platform is good enough to do real work, deployment feeds back into capability: more robots in the field generate more operational data, more production volume, and more process learning, which lowers cost and improves reliability, which expands the set of economically viable use cases, which pulls forward more deployment. Wright’s law then acts like a tailwind, if costs fall ~20% for each doubling of manufacturing volume, the winner will be the country that can double volume fastest and keep yield high while doing it.
If the U.S. does not own actuator throughput, it cannot ride this exponential. It will import the bottleneck module that gates volume, which means volume (and therefore learning) compounds inside the Chinese cluster instead of domestically.
The China Risk
There are three ways the strategy of dominating the NdFeB corridor can fail even if China wins the hardware race.
The Distribution Trap: If humanoids displace wage labor faster than household income mechanisms adjust, China gets more output capacity but weaker domestic demand. The automation dividend accrues to state-owned enterprises, leaving the core consumption problem unsolved.
The Protectionism Trap: The export strategy assumes foreign buyers will absorb the surplus. But “importing labor” is politically radioactive. If the West tariffs humanoids like EVs, China could be left with mass capacity and no end market.
The Bubble Risk: Local governments are already chasing this sector with subsidized capital. If incentives are misaligned, China risks building too many low-quality actuator plants, replacing a property/infastructure bubble with a robotics bubble.
The convergent outcome: a US–China contest for the NdFeB corridor
If the NdFeB → motor → actuator corridor is the gating module, and if that corridor is disproportionately Chinese, then China can do two things simultaneously: (1) scale embodied AI faster (because it controls throughput and allocation when markets tighten), and (2) price to clear (subsidize or underprice hardware to win share, then capture the profit pool upstream in the constrained layers). In other words, the robot export strategy is about locking the world into a dependency on the parts of the stack that actually determine scaling.
For the West, that creates a direct collision with politics. The U.S. and Europe will likely treat Chinese humanoids the way they increasingly treat EVs, batteries, drones, and telecom equipment: tariffs, local-content rules, procurement bans, security reviews, and “trusted supply chain” carve-outs. The more capable the robot, the more likely it triggers protectionism.
So the forward path is a classic two-step. China pushes to (i) dominate the corridor (NdFeB + actuator confirmed capacity, highest yield, fastest ramps), then (ii) export the surplus (finished robots and/or actuator modules) to keep its growth engine running.
The West responds by trying to (i) block the endpoint (tariffs/security), and then realizing it must also (ii) build the corridor (because you can’t regulate your way out of a physical bottleneck).
The Control Matrix
China has two primary levers to control this corridor: Export Controls (restricting volume) and Pricing (subsidizing costs). By toggling these two levers, Beijing can shift between strategic postures to thwart Western reshoring.
China can switch between them to exploit Western weaknesses:
If we game this out, China’s smartest play is not a blunt embargo (which unites the West), but a bifurcation of price.
China needs to keep moving up the value chain to escape its debt trap. The easiest way to achieve that is to engineer a persistent spread between the global price and the China price” of key inputs.
Inside China: Magnets are abundant and cheap (subsidized energy + lax environmental enforcement + VAT rebates).
Outside China: Magnets are scarce and expensive (due to export quotas, licensing friction, and no midstream).
This strategy turns Western capitalism against itself. Imagine you are a U.S. humanoid startup. Your software is world-class, but your Bill of Materials (BOM) is $5,000 higher than your Chinese competitor simply because you are buying magnets in Detroit instead of Dongguan.
Once you move the factory, the trap snaps shut. China captures the manufacturing value added, the supply chain ecosystem, and eventually the IP. The U.S. retains the “brand,” but the physical economy moves to China.
This is why the Blueprint below is not optional.
A tariff on Chinese magnets just makes the components more expensive for Western manufacturers, which actually accelerates the pressure to move operations offshore.
The only defense against a bifurcated price structure is to subsidize the spread. The U.S. must artificially lower the cost of domestic production until the Western price converges with the China price.
That brings us to the mechanism.
Break-the-cluster blueprint for the U.S.
Building a parallel loophole requires an industrial policy stack that is credible across many administrations. Separation, refining, alloy, magnets, and qualified actuator production are multi-year learning curves with long capex lead times and fragile early yields. If policy support is episodic, China can simply wait for the West to quit halfway through the ramp.
The U.S. cannot reach actuator throughput without first making the upstream corridor financeable, survivable, and continuous through price cycles.
Capital mobilization (build physical capacity, not just R&D).
The first layer is straight capex: fund the plants, equipment, and qualification processes across the corridor (oxide → metal/alloy → powder → magnet → motor/rotor → actuator). This is the hard part of industrial policy because it’s visible, expensive, and slow.
A corridor-wide price signal that beats China.
The Western failure mode is well known: you greenlight a multi-year build, then prices collapse before you reach stable yields, investors get wiped, and capacity dies. A price floor (or long-dated offtakes) is the antidote, but it has to be set with China in mind:
Floor price must be high enough to keep new Western capacity alive through the ramp
Effective protected price must survive a China underpricing cycle, which means the price floor + tariff (or equivalent border adjustment) has to clear above China’s export price.
Buffer stocks (time is the real scarce input in a crisis).
If actuators are the scaling unit, then a magnet/alloy buffer buys time for the rest of the corridor to keep running while capacity expands. You can’t scale a learning curve if you keep stopping the line.
Demand anchoring (guarantee throughput so Wright’s law can compound).
Even with capex and a floor, the corridor still needs volume to learn. This is where procurement and standard-setting actually matter: defense, logistics, critical infrastructure, and trusted industrial deployments can serve as the anchor demand that keeps lines hot and yields improving.
Conclusion: The Race for the Body
The next decade of geopolitical competition will be defined by a simple question: Can the U.S. build the body before China catches up on the brain?
The prevailing view in Washington is that the West can settle for a division of labor: we design the “Brain” (software/chips) while China builds the “Body” (actuators/magnets).
But this view is a strategic suicide pact.
It ignores the fact that China is rapidly scaling its domestic semiconductor value chain and competing effectively on software (a domain that is not materially constrained). China’s ability to iterate on foundation models and domesticate 7nm+ chip production is proving that the cognitive gap is closing faster than the West expects.
If China achieves parity on the brain while maintaining its chokehold on the body, the U.S. loses the entire platform.
China understands that in the physical world, throughput is leverage. By cornering the NdFeB-to-Actuator stack, Beijing is building a kill switch for the embodied AI economy. They are betting that while code has zero marginal cost, movement has infinite marginal cost if you run out of atoms.
The blueprint above (capital, price floors, and buffer stocks) is the only way to disable that switch. We cannot regulate our way out of a physical bottleneck, and we can no longer rely on a shrinking software lead to save us. If the U.S. wants to survive the age of physical AI, we have to build the body.
Another amazingly written post, thank you very much for your insights!
Amazing work, @CrossCurrents
Good to know that even if people debate the long-term volume assumptions, the core insight stands. If throughput is constrained, whoever controls the NdPr to magnet to actuator corridor controls the pace of scaling