Who are the Bad Players in Humanoid Robot Industry?

Official Alert: “150 Companies, One Product” — Real Risk Is Not Overcapacity, But Under-Innovation

On November 27, 2025, Li Chao, Deputy Director of the Policy Research Office at China’s National Development and Reform Commission (NDRC), issued a rare but precise warning to the booming humanoid robotics sector:

Over 150 companies are now active — more than half are startups or cross-industry entrants. While this fosters innovation, it also risks two critical problems:

(1) highly repetitive products flooding the market,

(2) compression of genuine R&D space.”

This is not a claim of “overcapacity.”
Global humanoid shipments in 2024 totaled only a few thousand units — firmly in the exploration phase.

The real danger?
A capital-fueled race to build the same demo — while avoiding the hard work of true innovation.

The Numbers: $14B in Funding, But Little Differentiation

• 2025 China humanoid robotics funding: >RMB 10 billion (~$1.37B)
• Number of companies: 150+
• Average funding per firm: <RMB 70 million (~$9.6M) — far below global peers
• Global comparison: Figure AI raised $1B in one round; two U.S. firms raised more than China’s entire sector combined

Yet despite the capital influx, product differentiation remains minimal:

• Most robots rely on off-the-shelf components from Chinese suppliers (Sanhua, Greenhorm, etc.)
• Core subsystems — planetary roller screws, force-torque sensors, high-precision harmonic drives — remain >80% imported
• Integration, not invention, defines most “R&D”

“The barrier to building a humanoid demo is now low,” said one investor.
“The barrier to building one that costs $20K and runs 24/7 in a factory? Still very high.”

A Familiar Pattern: The “Smart Compute” Mirage Repeats Itself

The humanoid boom mirrors China’s recent AI computing overbuild:

• Smart computing centers: Proliferating nationwide
• Average utilization: <30% (per Inspur AI Research)
• Result: Vast low-end capacity surplus, acute shortage of high-end, application-specific compute

In robotics, the same dynamic is emerging:

• Easy: Assemble a walking robot with standard actuators — ship a video, raise capital
• Hard: Develop domestic high-torque, low-backlash actuators; achieve sub-$20K BOM; secure 2,000-hour MTBF

Capital flows to the easy.
Progress requires the hard.

As the NDRC implies: when 150 firms chase the same low-hanging fruit, R&D bandwidth is diluted, and true breakthroughs starve.

Local Governments Join the Frenzy — With Identical Playbooks

At least 10+ provinces and 30+ cities have launched humanoid robot initiatives since 2024:

But global 2024 shipments were under 5,000 units — mostly for research, demos, and pilot deployments.

This disconnect — local ambition vs. global reality — risks a classic Chinese industrial overbuild:

Policy-driven, subsidy-fueled, and output-focused — without market pull.

The Corrective: Standards, Not Subsidies

On November 24, China’s Ministry of Industry and Information Technology (MIIT) took a constructive step:
It published the Humanoid Robotics Standardization Technical Committee — a 30-member body including:

• Leading builders: UBTECH, Unitree, Agibot, Robotera
• AI/software firms: Huawei, Xiaomi, SenseTime
• Industrial integrators: Chery, XPeng, Geely

This signals a shift:

From “build anything” → to “build what works, to shared standards.”

The goal is clear:
Let “good money” (serious R&D players) drive out “bad money” (just-to-get-funding players).

Who Are the “Bad Money” Players?

Three archetypes dominate the noise:

1. Concept-flip companies:
   • Former Floor Sweeping robots Companies
   • Add legs, rebrand as “humanoid leader”
   • Value: stock price lift, not product
2. Cross-industry opportunists:
   • Elder care, hotel, or real estate firms
   • Announce “robot joint ventures” with no tech depth
   • Use the narrative for policy access or PR
3. Pure integrators:
   • Buy Sanhua actuators, Greenhorm gears, NVIDIA Thor
   • Build a shell, claim “full self-research”
   • No core IP in mechanics, control theory, or AI

These players aren’t building robots.
They’re building valuation stories.

Investment Takeaway: The Bubble Is Real — But the Opportunity Is Realer

The NDRC isn’t killing the sector.
It’s pruning it.

For investors, the signal is clear:

• Avoid: Companies with only demo videos, no shipped units, no industrial customers
• Target: Firms with:
  • Vertical integration (actuators, control, AI)
  • Real B2B revenue (factories, logistics, hospitals)
  • Path to sub-$30K cost
  • Inclusion in MIIT standardization efforts

“We need more bubbles to attract talent and capital,” one industry veteran noted.
“But we need fewer companies pretending to be in the bubble.”