Humanoid Robot Market — $38 Billion by 2035
Goldman Sachs Research raised forecasts for the humanoid robot market to $38 billion by 2035 — six times the previous $6 billion estimate. Expected shipments reach 1.4 million units, with material costs dropping 40%. Main driver: AI progress exceeding expectations — robotic LLMs and VLA models help robots learn and adapt faster than anyone predicted.
But what does this mean for robotics engineers, especially those based in Vietnam? This article analyzes market dynamics, hiring trends, essential skills, and a concrete learning roadmap.
Market Landscape 2025-2026
China Leads Manufacturing
In 2025, China accounted for over 80% of all humanoid robots installed globally (~16,000 units per Counterpoint Research). Two major players:
- Unitree: 5,500+ units shipped in 2025 — more than Tesla, Figure, and Agility Robotics combined (each shipping only ~150 units). Unitree is preparing for an IPO on the Shanghai exchange targeting $610 million.
- AgiBot: 5,000+ units shipped, ranked by many analysts as the top shipper in 2025.
These numbers reveal a massive gap between China and the US in production scale. The US still leads in R&D and AI, but China now dominates manufacturing.
USA — R&D and Funding Leadership
- Boston Dynamics: Electric Atlas debuted at CES 2026; entire 2026 production committed to Hyundai and Google DeepMind. Price: ~$420K.
- Tesla: Optimus Gen 3 begins production summer 2026. Target price $30K — the cheapest on the market.
- Figure AI: $2.6B+ in investment, $40B valuation. Deploying Figure 02 at BMW factories.
- Agility Robotics: Digit robot for warehouse automation, partnering with Amazon.
- Apptronik: Apollo robot, partnering with Mercedes-Benz.
Other Players
- 1X Technologies (Norway): NEO — first humanoid designed for home use, $20K. Backed by OpenAI.
- Fourier Intelligence (China): GR-2 targeting healthcare and research, 53 DOF.
- UBTECH (China): Walker X, focused on enterprise and education sectors.
- Sanctuary AI (Canada): Phoenix, focused on cognitive AI and dexterous manipulation.
Hiring Trends 2026
Who Is Hiring Most?
Based on data from LinkedIn, Indeed, and specialized job boards (2026 Q1):
Unitree Robotics — strongest hiring in Hangzhou, China:
- Embedded engineers (motor control, real-time systems)
- RL researchers (locomotion, manipulation)
- ROS 2 developers
- Requirements: Mandarin Chinese is a significant advantage
Tesla Optimus — hiring in Palo Alto and Austin:
- ML engineers (end-to-end neural networks)
- Simulation engineers (Isaac Sim/Lab)
- Mechanical engineers (actuator design)
- Requirements: ML/DL background, Python + C++
Figure AI — hiring in Sunnyvale, CA:
- Robotics engineers (whole-body control)
- VLA researchers (vision-language-action)
- Perception engineers (3D vision)
- Requirements: PhD preferred, publication record
Boston Dynamics — hiring in Waltham, MA:
- Controls engineers (MPC, optimization)
- Perception/SLAM engineers
- Software engineers (C++, real-time systems)
- Requirements: Real hardware experience, strong C++
Agility Robotics — hiring in Corvallis, OR and Dallas, TX:
- Locomotion engineers
- Fleet management software engineers
- Requirements: ROS 2, Python, experience with legged robots
Compensation (2026, USA)
| Role | Entry (0-2 yr) | Mid (3-5 yr) | Senior (5+ yr) |
|---|---|---|---|
| Robotics Engineer | $120-150K | $150-200K | $200-300K |
| ML/RL Researcher | $130-170K | $180-250K | $250-400K |
| Controls Engineer | $110-140K | $140-190K | $190-280K |
| Embedded Engineer | $100-130K | $130-180K | $180-250K |
| Simulation Engineer | $120-150K | $150-200K | $200-280K |
Excludes RSU/stock — at startups like Figure AI, equity upside can be substantial.
Trends in China
In China, salaries are lower than the US but opportunities are more abundant:
- Unitree, AgiBot, Fourier, UBTECH are all hiring hundreds of engineers
- Salary range: 300K-800K CNY/year ($40-110K USD) for mid-level positions
- Advantages: Large-scale production, fast iteration cycles, more real robots to test with
Essential Skills
Based on analysis of 200+ humanoid robotics job postings (2026 Q1):
Tier 1: Must-Have
| Skill | Level | Where to Learn |
|---|---|---|
| Python | Proficient | Everywhere |
| C++ | Solid intermediate | LearnCpp, Effective C++ |
| Linear Algebra | Strong foundation | MIT 18.06, 3Blue1Brown |
| ROS 2 | Intermediate+ | ROS 2 Series on VnRobo |
| Reinforcement Learning | Understand concepts + can train | RL Basics, Spinning Up |
| MuJoCo or Isaac Sim | Hands-on experience | Simulation series |
| Git + Linux | Daily use | Pro Git, Linux Journey |
Tier 2: Highly Valuable
| Skill | Level | Where to Learn |
|---|---|---|
| Control theory | IK, PID, MPC basics | Humanoid Control |
| PyTorch | Train + deploy models | Official tutorials |
| Computer Vision | 3D perception, SLAM | SLAM Navigation |
| Imitation Learning | BC, ACT, Diffusion Policy | IL Series |
| Docker + CI/CD | Containerized development | Docker for Robotics |
| URDF/MJCF | Robot modeling | MuJoCo docs, ROS wiki |
Tier 3: Competitive Edge
| Skill | Level | Where to Learn |
|---|---|---|
| Whole-body MPC | Implement + tune | MPC Series |
| Sim-to-Real | Domain rand, deploy | Sim-to-Real Transfer |
| VLA models | Fine-tune, deploy | VLA Models |
| Hardware (actuators) | Understand motors, encoders, IMU | Datasheet reading, labs |
| Paper reading | Read and implement papers | arXiv, conference proceedings |
| Mandarin Chinese | Basic communication | Duolingo, HSK courses |
12-Month Learning Roadmap
Months 1-3: Foundations
- Python + C++ fundamentals: Build solid coding skills first if not already strong
- Linear Algebra + Dynamics: MIT OCW 18.06, An Introduction to Mechanics (Kleppner & Kolenkow)
- ROS 2 basics: ROS 2 Series — 4 articles
- MuJoCo setup: MuJoCo Tutorial — load model, simulate, visualize
Capstone Project: Control a robot arm in MuJoCo using inverse kinematics
Months 4-6: Robot Learning
- RL basics: RL Basics, OpenAI Spinning Up
- Humanoid-Gym: Part 4 of this series — train a walking policy
- Sim-to-Real concepts: Domain Randomization
- Imitation Learning: IL Series — Behavior Cloning, ACT
Capstone Project: Train a humanoid walking policy in Isaac Gym, verify transfer in MuJoCo
Months 7-9: Specialization
- Control theory: Humanoid Control — IK, operational space, QP
- MPC: Whole-Body MPC — iLQR, contact dynamics
- Loco-Manipulation: Part 5 — decoupled control, teleoperation
- Computer Vision: 3D perception, depth estimation for robotics
Capstone Project: Implement a whole-body QP controller for a humanoid in MuJoCo
Months 10-12: Portfolio and Applications
- Open-source contributions: Contribute to Humanoid-Gym, MuJoCo, or Isaac Lab
- Paper implementation: Choose 1-2 papers and implement them from scratch
- Portfolio: GitHub profile with 3-5 projects demonstrating depth
- Apply: LinkedIn, company websites, conference networking
Capstone Project: End-to-end demo — train RL policy → sim-to-sim verification → video demo
Opportunities for Vietnamese Engineers
The Reality
Currently, no company in Vietnam is building humanoid robots. But opportunities still exist:
- Remote work: Many US and European companies hire remotely for simulation, ML, and software roles
- China: Geographically closer, lower cost of living, many companies actively hiring
- Startup potential: Build a team in Vietnam, do R&D contract work for foreign companies
- Education: Teach robotics, create content, build community
Strengths of Vietnamese Engineers
- Cost efficiency: Companies can hire Vietnamese engineers at 30-50% of US costs for equivalent skills
- Strong technical skills: Vietnamese engineers are solid at coding, ML, and embedded systems
- Time zone: Overlap with both US (evenings) and China (daytime)
Challenges
- No real robots: Hard to practice on hardware — simulation only
- Networking: Far from major conferences (ICRA, IROS, CoRL)
- Visa: Difficult to obtain work visas for the US and Europe
How to Overcome These
- Simulation-first: Most research starts in simulation. MuJoCo and Isaac Lab are free.
- Open-source: Contribute code — get recognized in the community, build reputation
- Conferences: Submit papers, attend virtual sessions, or physical conferences held in Asia (CoRL, IROS when hosted in Asia)
- Buy a robot: Unitree G1 at $16K — expensive but feasible for a team or lab
- Community: Join the VnRobo community, share knowledge, learn together
Forecast 2027-2030
2027
- Tesla Optimus begins selling to external customers ($30-50K)
- Unitree IPO succeeds; production scales 10x
- Major universities begin establishing humanoid robot labs
2028-2029
- Humanoid price drops below $20K (entry-level tier)
- First truly general-purpose humanoid: capable of 10+ distinct household tasks
- Demand for robotics engineers increases 5-10x
2030
- Humanoids in factories become standard (like robot arms today)
- Home assistant humanoids begin commercial rollout
- The talent market shifts from "extremely rare" to "high volume demand"
Series Conclusion
Over 6 articles in the Humanoid Robot Engineering series, we have journeyed from:
- Platform Overview — Who is building what, and where the market stands
- Control Fundamentals — IK, operational space, balance control
- Whole-Body MPC — Real-time model-based control
- RL for Humanoids — Data-driven approach with Humanoid-Gym
- Loco-Manipulation — Combining walking and object handling
- The Future and Opportunities — Market, skills, career roadmap
Humanoid robotics is at an explosive inflection point — similar to deep learning in 2015 or autonomous driving in 2018. Engineers who start learning now will have a major competitive advantage in 3-5 years.
Where to start? Install MuJoCo, run Humanoid-Gym, and train your first walking policy. From there, everything will become clearer.
Related Articles
- Humanoid Robotics: Complete Guide — Full overview of humanoid robots
- Unitree vs Tesla Humanoid 2026 — Detailed comparison
- RL Basics for Robotics: From Markov to PPO — Getting started with reinforcement learning
- ROS 2 Introduction — Getting started with ROS 2
- Simulation for Robotics: Choosing the Right Simulator — Simulator comparison and selection guide
