Overview
The humanoid robot sector has attracted enormous investment since 2022, driven by the hypothesis that a general-purpose robot in human form can operate in existing human-designed environments without infrastructure modification. As of mid-2026, the gap between this hypothesis and demonstrated capability remains large — most commercial deployments are narrow in task scope, and the line between autonomous operation and teleoperation-assisted data collection is rarely disclosed clearly by companies.
Entries are split into US Companies and Non-US Companies. A standalone Overview & Critical Analysis entry addresses the autonomy verification problem, the specialized vs. humanoid debate, and investor sentiment. Valkyrie (NASA R5) documents the foundational government-funded platform whose actuator technology and engineering team directly seeded Apptronik.
Companies
US — Startups & Growth-Stage
| Company | HQ | Stage | Robot |
|---|---|---|---|
| Figure AI | Sunnyvale, CA | Growth / Pre-IPO | Figure 03 — general industrial |
| Agility Robotics | Corvallis, OR | Growth (Amazon-backed) | Digit — logistics tote handling |
| Apptronik | Austin, TX | Series A ($935M raised) | Apollo 2 (bipedal/wheeled, data-collection) — industrial; Apollo 3 commercial product targeted 2027 |
| 1X Technologies | Moss, Norway (US ops) | Series B | NEO — consumer/household |
| Sanctuary AI | Vancouver, BC | Series B | Phoenix — retail/commercial |
| Robust AI | San Jose, CA | Series B | Carter — logistics AMR (not humanoid) |
US — Public / Late Stage
| Ticker | Company | Notes |
|---|---|---|
| TSLA | Tesla | Optimus — internal R&D / learning mode, no external customers as of Q1 2026 |
| BDNSA | Boston Dynamics | Atlas — Hyundai automotive pilots; primarily R&D stage |
Non-US Companies
See Non-US Companies for full table. China dominates unit shipments (~90% of 2025 global volume).
Key Themes
- Autonomy gap: Most robots in 2025–2026 operate under partial or full teleoperation for non-trivial tasks. True closed-loop autonomy for novel physical tasks is a research problem, not a product reality.
- China’s manufacturing lead: Chinese firms shipped ~90% of global humanoid units in 2025. Price competition is extreme — Unitree R1 at $5,900 vs. US competitors at $100K+.
- Specialized robot alternative: For most defined industrial tasks, purpose-built robots (arms, AMRs, conveyors) outperform humanoids on cost, speed, and reliability. The humanoid value proposition is in unstructured, multi-task environments — a much harder problem than the current demos suggest.
- Battery life bottleneck: Agility Digit operates ~90 minutes before recharge. An 8-hour factory shift requires either swappable packs or docking infrastructure that undermines the “drop in anywhere” premise.
- Home robots move from promise to pre-order: 1X opened consumer pre-orders for NEO in October 2025 ($20,000 Early Access or $499/month), with first US home deliveries targeted for 2026 — the first humanoid a consumer can put a deposit on with a concrete delivery date, though ~30–40% of tasks still rely on remote “Expert Mode” teleoperation. Germany’s Neura Robotics (4NE1 Mini) and incumbents Tesla and Figure have also stated household ambitions, making the home market a genuine multi-player race rather than a single-company bet.
- Dexterous hands as the emerging bottleneck: With platform-level funding and battery/mobility problems increasingly well-trodden, industry attention is shifting to the hand/end-effector layer — see Robotic Hands & Dexterous Manipulation for the component-level suppliers (Inspire Robots, Tesollo, PSYONIC) racing to solve dexterity, tactile sensing, and field reliability.
- Data-collection infrastructure as a platform moat: Apptronik’s “Robot Park” and Neura’s “NEURA Gyms” both frame large-scale, real-world robot fleets as data-generation infrastructure for training foundation models (Google DeepMind’s Gemini Robotics; Neura’s Neuraverse), rather than treating deployed robots as finished commercial products — a structural pattern worth watching across the sector, not just at any one company.