Why the US–China Rivalry Runs Through a Silicon Chip

The contest is not over one invention or one factory. It is over a distributed industrial system in which design, machines, fabrication, packaging, materials and markets sit in different places.

026
CHIP
Strategic object
One chip · many borders

Its journey can cross American software, European machinery and Asian factories before reaching a Chinese or global device maker.

A leading-edge processor can train an artificial-intelligence model, guide a missile, optimise a warehouse or render a video game. The same underlying technology serves commercial and military systems. That dual use makes the semiconductor industry economically indispensable and politically sensitive.

The rivalry is often presented as a race between two national industries. The reality is more entangled. A chip is designed with specialised software, translated into masks, patterned with lithography, processed by hundreds of tools, cut, packaged, tested and assembled into a final product. No country independently dominates every stage.

The strategic power lies not in owning “the chip”, but in controlling a link that the rest of the chain cannot easily replace.

The supply chain is deliberately fragmented

As manufacturing became more complex, firms specialised. US companies became especially strong in design. Foundries concentrated capital-intensive fabrication elsewhere. Equipment, chemicals, wafers and packaging developed their own geographic clusters. A 2026 US Government Accountability Office report says roughly three-quarters of chips were manufactured and packaged in Asia as of 2022, while logic and memory production were particularly concentrated in Taiwan and South Korea.1

Value chain · Figure 1

Open the six linked stages

A restriction or disruption at one stage can propagate through the rest.

Design creates the blueprint

High value can be captured without owning the factory that makes the chip.

Geography turns specialisation into leverage

The United States has major strengths in chip design, design software and several categories of manufacturing equipment. The Netherlands hosts ASML, currently the only company able to produce extreme-ultraviolet lithography systems. Japan is important in equipment and materials. Taiwan and South Korea anchor advanced fabrication and memory. China combines a vast electronics market, assembly ecosystem, growing mature-node capacity and an expanding domestic technology base.

Industrial map · Figure 2

Activate the specialised clusters

USDesign + tools
NLEUV lithography
JPMaterials + tools
TWAdvanced foundry
KRMemory + foundry
CNMarket + manufacturing
One network, six clusters

Specialisation lowers duplication but creates strategic dependencies.

“Advanced” is only one part of the contest

Leading-edge logic and high-bandwidth memory matter for frontier AI and high-performance computing. Older process nodes remain essential to cars, factory equipment, power systems and household electronics. A shortage of inexpensive controllers can stop production just as effectively as a shortage of premium processors. China’s expansion in mature-node capacity therefore matters even when it does not reproduce the most advanced chip.

Two markets · Figure 3

Change the strategic lens

ADVANCED
AI · HPC
MATURE
cars · power
Advanced
Performance chokepoint

Control the chips and tools needed for frontier computing.

A lithography machine becomes foreign policy

Extreme-ultraviolet light enables critical layers in leading-edge manufacturing. ASML’s system fires laser pulses at tin droplets to create plasma, then guides 13.5-nanometre light through precision mirrors in a vacuum. ASML reported shipping 48 EUV systems in 2025.2 Because this capability is difficult to reproduce, access to lithography equipment is a strategic gate rather than an ordinary purchase.

Machine chokepoint · Figure 4

Turn on the EUV light path

Laser · tin · plasma · mirrors

The machine is a system assembled from specialised suppliers and decades of accumulated knowledge.

Export controls target capabilities, not all trade

US controls introduced since 2022 restrict specified advanced computing chips, high-bandwidth memory, semiconductor manufacturing equipment, software and support for certain end uses in China. The rules have repeatedly changed. In January 2026, the Bureau of Industry and Security moved specified processors such as Nvidia’s H200 and AMD’s MI325X to case-by-case licensing under security conditions—evidence that the policy is a managed gate, not a single permanent wall.3

Policy gate · Figure 5

Change the licensing posture

SUPPLIERchip · tool · software
CUSTOMERend user · end use · destination
Presumption or requirement

The transaction stops unless it meets the applicable rule and receives authorisation.

China treats dependence as a national vulnerability

China’s 2026 government work programme calls for decisive breakthroughs across the full integrated-circuit chain as part of greater technological self-reliance.4 Beijing also disputes the legitimacy and effects of US restrictions. In September 2025, the Ministry of Commerce opened an anti-discrimination investigation and argued that US measures harmed Chinese development interests and semiconductor supply-chain stability.5

Policy response · Figure 6

Two strategies alter the same network

UNITED STATESProtect and rebuild
  • Control selected advanced capabilities
  • Subsidise domestic fabrication
  • Coordinate with allied suppliers
  • Screen end users and diversion
CHINALocalise and substitute
  • Finance domestic chip ecosystems
  • Develop equipment and software
  • Expand manufacturing capacity
  • Use market scale to accelerate learning

Taiwan is central—but the map is moving

TSMC’s 2025 report says 74 percent of its wafer revenue came from processes at 7 nanometres and below. Its principal advanced manufacturing remains in Taiwan, while it is expanding production in Arizona, Japan and Germany.6 Governments and companies are paying to add geographic redundancy. Yet a new fab does not recreate the surrounding suppliers, workforce, utilities, tacit knowledge and customer relationships overnight.

Resilience dial · Figure 7

Move from efficiency toward redundancy

Lowest
unit cost
Duplicate
capacity
National
control
40%
Some resilience, continued interdependence

Redundancy reduces concentration risk but raises cost and still relies on foreign inputs.

Neither separation nor dependence is cost-free

The strongest inference from the supply chain is that complete technological separation would be slow, expensive and incomplete. Controls can delay access and redirect investment, but they can also encourage substitution, reduce supplier revenue and complicate allied trade. Subsidies can create capacity, but not instantly reproduce an ecosystem. Interdependence supports efficiency; concentration creates leverage and risk.

Knowledge check · Figure 8

What is the real strategic contest?

Choose the structural answer.
A semiconductor is the output of a network, not one national machine.
The chip rivalry is powerful precisely because the two economies are competing inside a system they still share.

Sources and further reading

Explainer 026 · World & CultureNext: Why A4 Paper Has Such a Strange Shape →