US-China Semiconductor Cold War Enters Its Most Consequential Phase

The third major package of US semiconductor export controls in eighteen months has shifted the contest from finished goods to the supporting infrastructure of the entire industry. The 2026 update extends restrictions to certain advanced lithography service contracts and broadens the foreign direct product rule. The intent is no longer to slow China's access; it is to fragment the global semiconductor supply chain along policy lines.

What changed

Three substantive items. First, the entity list has expanded to include several Chinese fabless designers and two domestic lithography aspirants. Second, the rules now restrict US-person services — meaning American engineers cannot legally service certain machines installed in covered fabs. Third, a more aggressive de minimis threshold catches more dual-use equipment.

According to Reuters reporting on industry briefings, the services restriction in particular is reshaping operating models at three of the largest equipment vendors. The compliance complexity is now substantial enough to affect time-to-revenue on individual tool sales.

What it means for allocators

The clean read is that capital intensity in the global semiconductor stack is going up, not down. Duplicate fabs, parallel supply chains, and redundant inventory all show up in capex lines. Margins in the equipment vendors compress before they recover, because the next leg of growth is funded by sovereign-backed customers building the parallel infrastructure.

This intersects with defense-tech capital flows. Sovereign-aligned semiconductor capacity is now treated as a national-security asset. The funding sources for the parallel infrastructure look more like defense procurement than commercial capex.

The Chinese response

Beijing's response has been measured publicly and aggressive operationally. State-directed lending into domestic equipment manufacturers has accelerated. Several Chinese fabs are now operating two-shift schedules to maximize tool utilization before parts shortages bite. The official line is that the country will achieve substitution in five years. The implicit acknowledgment is that the next three are going to be difficult.

Where the leverage actually sits

Despite the public framing, the leverage is asymmetric and uneven. The US has near-monopoly leverage in EDA software and advanced lithography. China has dominant leverage in critical mineral processing and certain mature-node capacity. Neither side can fully decouple without absorbing costs that voters and party leadership will struggle to accept.

Per analysis published by Bloomberg, the realistic outcome is a partial decoupling at the bleeding edge, with mature-node commerce continuing largely uninterrupted. That bifurcation is itself a policy outcome neither capital nor the chip companies have fully priced.

What to watch

Two markers. First, the next round of US person services guidance — clarifications matter as much as the rule itself. Second, whether Chinese SMIC achieves credible 5-nanometer yield at commercial scale. Both will resolve significant uncertainty inside the next eighteen months.