NdFeB Magnets for US Humanoid Robotics: Figure, Tesla Optimus, Agility Digit, and the 2026 Supply Chain Reality

The US humanoid robotics industry scaled faster in the last 12 months than any permanent-magnet motor segment since the EV boom started. Tesla commenced Optimus Gen 3 production at its Fremont facility in March 2026 with a stated capacity target of 1 million units per year — a figure that, if achieved, would consume meaningful percentages of global NdFeB output on Optimus alone. Figure AI's BotQ facility in San Jose is tooled for 12,000 Figure 03 units per year, with BMW running a pilot deployment for material handling and parts transfer. Agility Robotics' Digit has moved past pilot into operational use across Amazon fulfillment centers, Spanx distribution, and GXO logistics. Apptronik's Apollo is in NASA collaboration; 1X Technologies' NEO is shipping to enterprise pilots. The question for 2026 is no longer whether US humanoid robots work in real environments. It is whether the magnet supply chain can support their production ramp.

A typical humanoid robot contains 20–50 actuator motors depending on degrees of freedom. A bipedal platform like Agility Digit has around 16 actuated joints. A full humanoid like Figure 03 or Optimus Gen 3 has significantly more — Tesla's February 2026 reveal showed Gen 3 hands alone using 50 actuators per hand, which means a complete Optimus Gen 3 contains well over 100 magnetic motors. The high-torque joints (hip, knee, shoulder) use radial multi-pole sintered NdFeB rings as the rotor, typically 40–120mm outer diameter with 8–12 poles. Lower-torque joints (wrist, fingers, ankles) use smaller rings or coreless motor configurations. The grade choice is driven by continuous operating temperature — joint motors under sustained load routinely reach 100–130°C internal temperature, which puts N42SH (150°C rated) as the floor and N45SH or N48SH as the working standard.

• •Hip and knee joints: 360 N·m peak torque class — requires N48SH or N45UH radial multi-pole rings, 60–120mm OD
• •Shoulder and elbow joints: 120–220 N·m — N45SH rings, 40–80mm OD
• •Wrist joints: 30–60 N·m — N45SH or N42SH compact rings, 20–40mm OD
• •Finger actuators: ≤10 N·m — miniature coreless or small ring assemblies, N42 to N45SH depending on duty
• •Total NdFeB content per humanoid: 1.5–3.5 kg depending on platform and generation

MP Materials shipped its first commercial NdFeB magnets from the Independence Facility in Fort Worth, Texas in December 2025 — the first commercial sintered NdFeB production in the United States in over two decades. Capacity is planned at ~1,000 metric tonnes/year at Independence, with the 10X Facility in Northlake, Texas breaking ground in 2026 and targeting first commercial product in 2028. Combined, Lynas, MP Materials, Quadrant and smaller Western sources currently represent under 5% of global sintered NdFeB production. Total 2027 humanoid robot magnet demand alone, if current ramp trajectories are realized, is forecast to approach 2,000–4,000 tonnes — a volume that Western capacity cannot meet on its own for the remainder of the decade. This is not a hypothetical supply constraint. It is the actual operating environment US robotics companies are navigating today.

China's April 2025 MOFCOM export controls on heavy rare earth elements continue to apply to SH, UH, and EH grade NdFeB magnets shipped to US destinations. The October 2025 expansion was suspended from November 7, 2025 through November 10, 2026, but the underlying April 2025 rules remain fully enforced. In practice, commercial US humanoid robotics programs have seen routine MOFCOM approval when documentation is correct — end-use as commercial robotics, named US end-customer, clear production volume and timeline. Rejections are concentrated in defense-adjacent programs and applications with unclear civilian justification. US robotics procurement teams should budget a 45–60 day license timeline in addition to production lead time, and qualify suppliers that operate their own MOFCOM licensing capability rather than routing through trading companies that add delay and opacity.

Specifying magnets for humanoid robot joint motors comes down to three parameters: peak continuous operating temperature, required torque density, and coercivity margin against fault conditions. The decision tree most US humanoid programs follow:

• •Max joint internal temperature ≤120°C (cool-running, well-cooled design): N45H or N48H acceptable — lower cost, adequate thermal margin
• •Max joint internal temperature 120–150°C (typical sustained walking/load): N42SH or N45SH — the working standard for mainstream humanoid joint motors
• •Max joint internal temperature 150°C+ or premium performance programs: N48SH or N45UH — required for high-torque, high-duty-cycle designs
• •Always specify minimum HcJ at working temperature, not at 20°C — thermal derating of coercivity is the dominant design concern
• •Request GBD-processed variants — reduces Dy/Tb content by 50–70% with equal performance, reducing both cost and supply-chain risk

The qualification criteria that actually matter for humanoid robotics are narrower than the generic automotive checklist. Torque smoothness is non-negotiable — radial multi-pole ring capability is mandatory, not optional. Dimensional precision is automotive-class or tighter — ID tolerance ±0.02mm, concentricity ≤0.02mm. Surface flux consistency must be <1% across a production lot, and 100% flux mapping with serial-number traceability is a reasonable request from serious buyers. Beyond technical qualification, operational qualification matters: can the supplier run 50-piece prototype lots in 2–3 weeks, can they deliver PPAP documentation on request, and do they operate their own MOFCOM export licensing. The last point is often the deciding factor — a supplier without in-house licensing adds weeks of unpredictable delay to every shipment.

• •Radial multi-pole ring capability: verified with production samples, not claimed in marketing
• •Tolerance capability: request CPK data on ID and concentricity from recent production
• •Flux consistency: batch-level flux mapping report, not 'typical' values
• •Prototyping speed: 50–100 piece lots delivered in 2–3 weeks is the qualifying standard
• •MOFCOM licensing: in-house, not through a trading intermediary
• •IATF 16949: reasonable baseline even though robotics is not automotive — indicates process discipline

Every serious US humanoid robotics procurement strategy in 2026 is dual-sourced. The practical structure that has emerged in the market: 80–90% of volume from qualified Chinese suppliers (for cost, grade availability, and capacity), 10–20% strategic volume from Western sources (MP Materials for US, Lynas/Australia for some programs) as contingency and for customer programs requiring domestic content. This is not political posturing — it is an operational response to the supply environment. Chinese suppliers deliver SH and UH grade rings at scale, with GBD processing, at prices Western producers cannot match at current capacity. Western producers are scaling but will not reach humanoid-scale volume until 2028 at earliest. The dual-source structure keeps programs moving during any future licensing disruption without demanding impossible cost trade-offs.

We supply magnets to robotics programs across the United States, Europe, and Asia, including humanoid, collaborative, and industrial robotic platforms. Our radial multi-pole ring capability covers 4 to 14 pole configurations with sinusoidal surface field output, GBD-processed SH and UH grades at competitive cost, and full flux mapping with per-piece traceability. We operate in-house MOFCOM export licensing with a documented track record of commercial-robotics approvals for US destinations. Prototype batches of 50+ pieces ship in 2–3 weeks; production qualification follows on typical Tier 1 timelines.

• •Radial multi-pole rings in 4–14 pole configurations, OD 15–250mm
• •GBD-processed N42SH through N48UH at production volumes
• •100% flux mapping with serial-number traceability
• •Dimensional tolerance to ±0.02mm on ID, ≤0.02mm concentricity
• •In-house MOFCOM export licensing — no trading company intermediary
• •IATF 16949 certified quality management system
• •Dedicated engineering support in English, UK office for EU/US buyers