China’s Chip-Making Capacity and the Limits of Containment

China can now mass-produce 7 nanometer chips without Western lithography, making it a formidable fast follower but not a frontier peer. That single fact anchors everything else in the semiconductor contest, from export controls to the global spread of Chinese AI models.

*This report deliberately excludes Taiwan. It covers only fabrication capacity on the Chinese mainland, where the policy questions for Washington and investors actually sit.

What the Mainland Can Actually Make

SMIC, the national champion foundry, mass-produces at 7 nanometers using deep ultraviolet (DUV) machines pushed far beyond their design intent. The workaround is expensive. Reaching 7 nanometers with DUV requires roughly 34 lithography steps where extreme ultraviolet (EUV) tools need nine. Costs run an estimated 40 to 50 percent above comparable foundry processes, and yields remain well below industry norms.

Capacity is nonetheless scaling fast. SMIC’s output at 7 nanometers and below reached an estimated 45,000 wafer starts per month in late 2025. That figure is projected to hit 60,000 this year and 80,000 in 2027, with 7 nanometer capacity alone set to double in 2026. Beijing wants combined 7 and 5 nanometer output to grow fivefold within two years, with Hua Hong joining SMIC at the advanced edge.

A 5 nanometer process exists, but only barely. Pilot runs are underway for Huawei’s next Ascend accelerators. Independent estimates put yields near 20 percent, though some Chinese sources claim figures two or three times higher.

Below the leading edge the picture is far stronger. China commands enormous and growing capacity at 28 nanometers and above, where most automotive, industrial, and consumer silicon lives. Memory is following the same trajectory, with YMTC and CXMT scaling output and CXMT beginning high-bandwidth memory development.

The practical ceiling is this: domestic AI chips fabbed in 2026 roughly match Nvidia’s 2022 generation. That is sufficient for mass domestic inference and a captive market. It is not sufficient for frontier training at competitive cost.

The Vertical Bet

China’s industrial policy has gone fully vertical, funding every layer from photoresist to processor. Big Fund III, a $47.5 billion state vehicle that began deploying in 2025, prioritizes the weakest links: lithography, fab equipment, and design software.

The 15th Five-Year Plan, covering 2026 to 2030, formalizes five priorities: