Humanoid robotics supply chain in 2026: sourcing for Optimus, Figure, 1X, Apptronik, Agility, and the rest

Humanoid robotics moved from research project to production category in 2024 and 2025. Tesla Optimus, Figure 02 and 03, 1X NEO, Apptronik Apollo, Sanctuary AI Phoenix, Agility Robotics Digit, and Unitree H1 and G1 are all in customer pilot or early commercial deployment. The supply chain that needs to feed this wave looks more like consumer electronics scale combined with industrial robotics qualification, and the existing supplier base is not yet sized for it. Lean SupplAI was built to track humanoid-robotics-specific suppliers alongside the broader robotic components index, because the qualification questions and capacity constraints are different from traditional industrial robots.

For procurement teams sourcing for humanoid programs in 2026, the binding constraint is rarely the obvious component. Custom actuators, harmonic drives at unusual sizes, and tactile sensors are the categories where supply lags demand most acutely.

The named programs

Tesla Optimus targets in-house production at Fremont, with Tesla buying components rather than full assemblies. Figure (Figure AI) builds Figure 02 and 03 with a hybrid in-house-and-partner model, and recently announced BMW partnership for plant deployment. 1X Technologies builds NEO with humanoid-specific actuators developed in-house. Apptronik Apollo emerged from human-cooperative robotics research and partners with NASA. Agility Robotics Digit is targeting warehouse logistics with Amazon and GXO as early customers. Sanctuary AI Phoenix targets general manipulation. Unitree H1 and G1 are the fastest-iterating Chinese options.

Actuators: the binding constraint

Humanoid actuators are typically smaller, denser, and higher torque-per-mass than industrial actuators. The dominant in-house and merchant actuator suppliers include Maxon (compact servo), Faulhaber, Harmonic Drive LLC (strain-wave gearboxes at humanoid sizes), Nabtesco (cycloidal), and a growing tier of Chinese suppliers (Leaderdrive, ZeroErr Control). 1X and Tesla build their own actuators. Allocation on harmonic drives at the sizes humanoids need (CSF-14 to CSF-32 range) has been tight through 2025.

Lithium batteries and power management

Humanoid power systems lean on high-energy-density lithium chemistries. LG Energy Solution, Samsung SDI, CATL, and Murata supply cells. Custom battery packs typically come from BYD, EVE Energy, or smaller specialty pack assemblers. For motor drivers, ODrive Robotics, Elmo Motion Control, ESCON (Maxon), and increasingly humanoid-specific drivers from Tesla and 1X internally are the qualified options.

Sensing and AI compute

For depth perception: Intel RealSense, Orbbec, Zivid, plus increasingly LiDAR from RoboSense and Hesai. For force-torque sensing: ATI Industrial Automation, Robotous, and OnRobot are the qualified industrial options. For tactile: Tekscan, Digit (CMU spinout), and increasingly in-house for the leading programs. For AI compute on the robot itself: NVIDIA Jetson Orin and Thor are the dominant choices, with Qualcomm RB6 and AMD Versal as alternatives. Most humanoid programs run hybrid local-and-cloud inference.

Where the supply gaps are in 2026

The most acute supply gaps in 2026 are: small-form-factor harmonic drives, high-density humanoid-grade actuators, robust tactile sensors, and custom cable assemblies for the high cycle counts humanoids generate. Most major programs have addressed these by building in-house, which means the merchant supply base is sized for second-tier programs rather than scale leaders.

How Lean SupplAI tracks the humanoid stack

Lean SupplAI indexes humanoid-relevant suppliers across actuators, harmonic drives, motor controllers, batteries, sensing, and on-robot AI compute, with humanoid-specific qualification attributes (cycle counts, density, weight, integration footprint) tracked alongside the standard industrial-robotics attributes. For procurement teams running humanoid programs, Lean SupplAI surfaces the supplier candidates that actually match the use case, with allocation status visible inline.

What sets Lean SupplAI apart