Key insights
- A report indicates China is gaining dominance in robotics due to its manufacturing scale and state-led industrial policy, while the US leads in AI and semiconductors. China's higher robot deployment and state-funded training farms give it an advantage in real-world AI learning. This trend suggests a potential drag on US equity valuations in sectors reliant on automation and robotics, as China captures more of the economic value.
Investing.com -- The global race for artificial intelligence supremacy is increasingly being determined by industrial scale rather than compute power alone, according to an Alpine Macro report published on May 12, 2026.
While the United States maintains a clear advantage in frontier intelligence and silicon development, China has established a dominant position in the physical manufacturing, supply chain infrastructure, and deployment density required for robotics.
The report indicates that a bifurcated hardware landscape has emerged.
The U.S. continues to lead the "brain layer" through advanced semiconductors and software autonomy, while China commands the "body layer" by utilizing highly integrated manufacturing clusters to compound advantages in cost, data collection, and production across both humanoid and industrial robots.
Under its 15th Five-Year Plan, Beijing has explicitly designated "embodied intelligence" as a core pillar of national development, deploying billions of dollars in specialized economic development funds.
As noted by Alpine Macro Chief Geopolitical Strategist Dan Alamariu, "State capitalism is needed to ’win’ at AI. AI success depends on forms of heavy state interventionism, a departure from techno-libertarian dreams."
The current state-led ecosystem has widened the physical deployment gap between the two nations.
According to the International Federation of Robotics, China leads the world with more than 295,000 annual industrial robot installations, compared to 34,200 units in the United States.
The above figures directly impact physical AI learning curves, which rely heavily on real-world operational hours rather than digital data repositories.
China has industrialized training systems via state-funded "robot training farms", such as the Wuhan Hubei Humanoid Robot Innovation Center, which generates approximately 100 hours of usable embodied data per day.
The country’s manufacturing density allows leading Chinese producers like Unitree to aggressively lower pricing.
Unitree scaled humanoid sales from 5 units in 2023 to 3,551 units in the first nine months of 2025 as its average unit price fell from roughly RMB 593,000 to RMB 168,000.
Conversely, the U.S. robotics model focuses on frontier-grade silicon embedded directly on physical units, leveraging advanced compute stacks to generate synthetic training data through high-fidelity simulations like NVIDIA’s Cosmos world models.
However, the U.S. remains structurally dependent on Asian manufacturing channels.
China currently supplies an estimated 80% to 90% of key hardware components globally, controls 93% of the permanent magnet market essential for robot actuators, and manages nearly 99% of heavy rare earth element processing required for high-strength robotic motors.