U.S. Approves TSMC Chip Shipments to China for 2026

With a major export privilege expiring on Dec. 31, the U.S. issued TSMC an annual license to continue importing U.S.-made semiconductor manufacturing equipment into China's Nanjing fab, according to Reuters. This license "ensures uninterrupted fab operations and product deliveries," TSMC said.

The critical shift is not that the factory remains operational. It’s how the factory remains operational.

The old system worked like VIP access. Approved sites could keep receiving certain controlled materials without repeated paperwork hurdles. The new system looks like visitor access: allowed, but renewed on a stricter timeline, starting with a one-year license for 2026.

TSMC said this annual license allows export-controlled goods to be shipped to its Nanjing plant without requiring individual vendor approvals every time. Other tech giants like Samsung Electronics and SK hynix also received similar licenses.

This is important, even though TSMC does not produce its most advanced semiconductors in Nanjing. The facility produces 16-nanometer and older-generation chips, the kind used in vehicles, routers, consumer electronics, and industrial machinery. When such components face a squeeze, it manifests as delayed shipments, rising prices, and longer repair timelines.

Context from TSMC’s own reporting keeps this in perspective. In its 2024 annual report, the Nanjing site accounted for 2.4% of revenue. It is not a core business, but it is significant enough to cause disruption.

The Permission Economy

This annual-license model signals a broader shift: chip supply chains are moving toward renewable approvals that regulators can tighten or loosen without rewriting the entire rulebook.

But it doesn't end with factory tools. The same permission logic is already appearing in AI hardware flows. Nvidia, for instance, has asked TSMC to ramp up production of H200 AI chips, with additional output expected to begin in Q2 2026, according to Reuters. Chinese firms have placed orders for more than 2 million H200 chips for 2026, while Nvidia currently has supply for approximately 700,000, according to sources.

In this sector as well, approval gates remain critical. China has not yet approved shipments of H200 chips, leaving uncertainty around how quickly those orders will translate into deliveries.

The implication is clear: the movement of tools, wafers, and AI accelerators is not limited to engineering and demand anymore. It happens within a growing web of permissions, which are becoming more time-bound and actively managed.