The US MATCH Act Was Built for a Different Architecture

A chokepoint designed for one architecture is being answered from another. The proposed US MATCH Act targets deep-ultraviolet lithography access because the assumption baked into the bill is that horizontal node scaling remains the binding constraint for memory density. 3D DRAM breaks that assumption. Vertical stacking means bit density can keep rising through height, not width, and height can be cut with multi-patterning and design-technology co-optimization on tools China can now build.

Shanghai Yuliangsheng Technology is producing 28nm-class ArF immersion DUV machines, with five units shipping this year and roughly twenty next, and Beijing has put CXMT, SMIC, and Hua Hong at the front of the delivery line. Most parts are domestic; critical components still come from Japan. The line is substitution, not catch-up: 28nm immersion is generations behind the leading edge.

CXMT's bet is the architecture shift. Its vertical GAA wordline plus horizontal capacitor DRAM cell, wafer-to-wafer hybrid bonding at the Hefei pilot, and the multi-patterning and DTCO stack turn the old chokepoint into a road that bends. The MATCH Act was written for a world where density runs in two dimensions. That world is closing.