China supplies 88.6% of EU external magnet imports
AIR 005 · China–Europe Trade Conflict
Independent, self-funded research. Trade: CEPII BACI; production: OECD ICIO; policy: primary notices.Independent, self-funded research.
ContactChina's share of the $72.5 billion trade in robot, sensor, control and motor hardware has more than halved since 2010, and half of the trade sits in controls and power.
Physical AI Product Network
Mexico is now the largest origin of U.S. physical-AI hardware; China is second
Exhibit 1. U.S. imports by origin, 2024, US dollars and share of the total
Note: “Other Asia, n.e.s.” is the reporting code commonly used for Taiwan.
Source: CEPII BACI HS02 (V202601); Symbiosis United Analytics analysis.
Physical-AI systems (robots, autonomous vehicles, automated logistics) depend on a hardware stack well beyond the robot itself: cameras, navigation devices, sensors, control boards, power converters and motors.
This report screens U.S. imports of that stack across six hardware functions from 2005 to 2024, measures how much comes from China, and traces China-origin exposure into U.S. production using input-output tables.
Controls and power alone account for $37.4 billion of the $72.5 billion total. China supplies 11.2 percent of that function, and between 10.2 and 15.3 percent of every other function except finished robots (3.6 percent).
Half of U.S. physical-AI hardware imports are controls and power; China supplies one-eighth overall
Exhibit 2. U.S. imports by hardware function, 2024, US dollars, with China-origin share
Note: Hardware functions group the HS6 products in the AIR 004 screen.
Source: CEPII BACI HS02 (V202601); Symbiosis United Analytics analysis.
China's share of U.S. imports peaked at 28.8 percent in 2010, stayed between 25.2 and 27.2 percent through 2018, and has fallen every year since, to 12.5 percent in 2024. The decline is not only relative: China-origin imports peaked at $14.6 billion in 2018 and were $9.1 billion in 2024, while total imports kept growing.
China's share of U.S. physical-AI hardware imports has more than halved since its 2010 peak
Exhibit 3. China-origin share of U.S. imports of the screened products, 2005–2024
Source: CEPII BACI HS02 (V202601); Symbiosis United Analytics analysis.
Mexico supplied $16.3 billion in 2024, 22.4 percent of U.S. imports and 1.8 times China's $9.1 billion (Exhibit 1). Germany, Japan, Canada, Vietnam and Thailand each supply between 5.7 and 7.5 percent.
Input-output modelling allocates the $9.1 billion of China-origin imports to the U.S. sectors that use them and traces the linkages onward. The resulting footprint is $13.0 billion across U.S. production, 1.44 times the direct flow, with $6.1 billion of associated value added. Machinery and motor vehicles carry the largest footprints.
Machinery and motor vehicles carry the largest modelled footprint of China-origin hardware
Exhibit 4. Modelled downstream footprint by U.S. production sector, eight largest, US dollars
Note: Modelled with OECD input-output tables; a screen for diligence, not a forecast.
Source: OECD ICIO, 2023 edition (2020 table); CEPII BACI HS02 (V202601); Symbiosis United Analytics analysis.
Cameras, navigation devices, sensors, controls, converters and motors determine whether a system can perceive, decide and move.
Alternative origins matter only after qualification, safety requirements, firmware, calibration and customer approvals are considered.
The next diligence step is a bill of materials and spares register organised by perceive, decide, control, move and power functions.
This report is part of our independent public research. It was self-funded and was not commissioned or funded by any company, government or client. Edition 1.0, published 21 June 2026.
Symbiosis United Analytics (2026). Physical AI Product Network. Actionable Insight Report 004, edition 1.0. https://symbiosisunited.com/research/air-004/
@techreport{sua2026air004,
author = {{Symbiosis United Analytics}},
title = {Physical AI Product Network},
institution = {Symbiosis United Analytics},
type = {Actionable Insight Report},
number = {004},
note = {Edition 1.0},
year = {2026},
month = jun,
url = {https://symbiosisunited.com/research/air-004/}
}
We apply the methods behind this research to an organisation's own question, from an exposure screen to a decision memo. How we work with clients, or get in touch.