While US Shows Demo Videos, Chinese Robots Enter Warehouse POCs

Why do US robotics companies, valued in the billions, keep releasing "Coming Soon" demo videos, while Chinese robotics companies have already signed batch delivery contracts and even started calculating after-sales maintenance costs?

This question was recently publicly highlighted by Oren Etzioni, founding CEO of the Allen Institute for AI. He directly said: Musk promised humanoid robots, China delivered them. This observation stings, but for entrepreneurs who have done actual projects, it's not surprising at all.

I started working on warehouse robots in 2019, stepping on countless pitfalls from SLAM mapping to motion planning. At the time, a saying was popular in the industry: US companies build "general platforms," Chinese companies build "scenario customizations." Three years later, the result is obvious.

Two Routes, Two Survival Logics

US Route: High Valuation, Long Cycle, Capital Dependent

Representative companies include Tesla Optimus, Agility Robotics, Boston Dynamics. Their characteristics are: cool hardware design, top-tier motion control algorithms, but at least one "supply chain hell" away from mass production. Boston Dynamics' Atlas can do backflips, but each unit reportedly costs over $2 million. Agility's Digit bipedal robot announced commercialization in 2023, but public batch orders remain scarce to this day.

China Route: Low Cost, Fast Iteration, Run First

Representative companies include Unitree Robotics, Fourier Intelligence, Agibot. Unitree's H1 humanoid robot achieved small-batch delivery as early as 2024, with prices as low as around $90,000. Fourier's GR-2 has been implemented in rehabilitation and industrial scenarios. Agibot is more direct, launching the "Expedition A2" in August 2024, with the goal of entering factories to work.

Note that this is not a technology gap, but a fundamental difference in business models. Chinese robotics companies recognized from the start: if a humanoid robot cannot help customers save costs within a year, customers won't buy.

Core of Technical Implementation: Cost and Deployment Efficiency

I deployed two AGVs in a warehouse, one using a SLAM solution from a well-known US company, the other self-developed by a domestic team. For the same mapping task, the US solution required professional engineers to spend 3 days tuning parameters, while the domestic solution used adaptive algorithms and finished in 1 day. It's not about who is smarter, but who is closer to the field.

# Comparison of SLAM parameters in actual warehouse deployment (simplified example)
# US Solution: Approx. 15 key parameters requiring manual adjustment
slam_params_usa = {
    'scan_matcher': 'ICP',
    'loop_closure_threshold': 0.6,
    'motion_model': 'constant_velocity',
    'optimizer': 'g2o',
    'max_iterations': 500,
    'min_linear_distance': 0.5,
    'min_angular_distance': 0.3,
    # And more...
}

# China Solution: Adaptive parameters, only need to set map resolution
slam_params_cn = {
    'resolution': 0.05,
    'mode': 'adaptive',
    'max_workers': 4
}

Have you calculated deployment costs? The US solution typically requires purchasing additional industrial PCs, LiDAR, etc., raising hardware costs by 30%. The China solution is often integrated into the robot body itself, sometimes running on low-cost ARM chips.

[!note] Key Conclusion

The biggest gap between US and Chinese robotics companies is not technology, but "productization" and "cost control" capabilities. Chinese companies are better at turning lab technology into deliverable, profitable products.

Entrepreneur's Perspective: Cash Flow Matters More Than Valuation

I've seen too many teams whose PPTs say "General Purpose Humanoid Robot," raising tens of millions, yet unable to produce a stable gait. Meanwhile, my brothers doing warehouse robots focused from the start on ROI calculations of "unit cost vs. labor replacement cost."

For example, a warehouse handling robot sells for 80,000 RMB, can replace 1.5 workers, and pays for itself in one year. This is the hard metric customers are willing to buy. Chinese companies generally achieve this because their supply chains are mature, and they are willing to iterate quickly at low margins.

The problem with US companies is: they are accustomed to the valuation logic of "selling dreams," have a strong engineering culture, but extremely high operating costs. Agility's Digit reportedly sells for $250,000; customers might as well use traditional AGVs plus robotic arms. While Unitree pushed the price down to $90,000 and provided 7x24 remote operations and maintenance.

Image Description

This image is CCTV coverage of a Chinese robot production line, with robots densely packed on the assembly line being assembled. This is the confidence of Chinese robotics companies: not relying on a few engineers tuning parameters in a lab, but relying on engineers and production line workers in the factory to build and sell robots one by one.

Advice for Entrepreneurs

If you want to enter embodied intelligence now, I suggest:

• Choose a vertical scenario first (warehouse, housekeeping, inspection), don't go general purpose right away
• Control

Original Link: https://www.tmtpost.com/8064381.html