SIA and BCG: by 2032, the US will have the capacity to manufacture 28% of chips below 10nm, while China is expected to make only 2% of the most advanced chips
Nikkei Asia :
Context & Ripple Effects
The forecast extends an earlier SIA-BCG projection that the US would triple domestic chip-manufacturing capacity by 2032, separating growth in overall fabrication from its position in sub-10nm capacity.
It frames the advanced-node contest as distinct from China’s broader self-sufficiency push: China’s domestic chip supply has been projected to rise even while its share of the most advanced production remains limited.
First-order effects
- The projection positions the US for a substantially larger share of sub-10nm manufacturing capacity by 2032, strengthening the domestic supply base available to companies that need leading-edge chips.
- China is projected to account for only 2% of the most advanced-chip output, leaving its advanced-chip supply more constrained than its broader domestic production ambitions.
Second-order effects
- A wider US-China gap at advanced nodes raises the value of alternative performance routes, including advanced chip packaging, where both countries have sought to improve system performance.
- Chip buyers and infrastructure planners may place greater weight on where leading-edge capacity is located, rather than treating aggregate domestic chip-output growth as a proxy for access to the highest-end parts.
Third-order effects
- If these capacity projections materialize, semiconductor supply chains will become more segmented by technology tier: mature-node self-sufficiency can grow independently of access to leading-edge manufacturing.
- The pattern reinforces compute capacity as strategic leverage, with national industrial policy increasingly judged by its ability to support advanced production rather than fab investment alone.
The trend: Advanced semiconductor policy is shifting from expanding domestic chip output in general to securing capacity at the leading technology nodes that underpin high-performance computing.