New US AI Chip Export Levy Looms

This policy is not a simple tax but a multi-faceted strategy combining tariffs and unprecedented revenue-sharing deals. In January 2026, a Section 232 proclamation imposed a 25% tariff on advanced AI chips imported into the U.S. specifically for re-export to China. This forces manufacturers to route China-bound chips through the U.S. for testing and certification, applying a significant cost to the export process. In a separate, more direct arrangement, companies like Nvidia and AMD secured export licenses for their specifically designed, less powerful chips for the Chinese market, such as the H20 and MI308. The condition for these licenses is an agreement to pay the U.S. Commerce Department 15% of the revenue from these sales, a move confirmed by President Trump after negotiations with Nvidia CEO Jensen Huang. This revenue-sharing model represents a significant policy shift, moving from outright bans to a monetization of market access. The U.S. Commerce Department could collect an estimated $2 to $3 billion annually from this arrangement. For Nvidia alone, based on past earnings, this could mean payments of up to $2.7 billion per year to the U.S. government for continued access to the Chinese market. The policy follows a series of tightening export controls that began in October 2022, aimed at restricting China's access to high-performance GPUs like Nvidia's A100 and H100. These earlier bans created significant financial headwinds; Nvidia, for instance, reported a potential $5.5 billion inventory write-down due to the restrictions on its China-focused chips. Navigating these shifting regulations creates a complex compliance landscape for manufacturers. Companies must now design export-compliant chips with reduced performance to meet specific thresholds, a costly process. The high price of non-compliance was recently highlighted when one semiconductor equipment firm was fined over $252 million for illegal re-exports to China. The high fixed costs of semiconductor development—with design for a 5nm chip costing around $540 million and factory construction running into billions—make global market access critical for funding next-generation R&D. One analysis projected that a 50% decoupling from the Chinese market could cost U.S. chipmakers over $41 billion in annual revenue and cut R&D funding by roughly $8 billion. This "pay-to-play" model has raised legal questions, with some analysts suggesting it may violate the U.S. Constitution's prohibition on export taxes. The strategy signals a move from pure technological containment to a more transactional approach, balancing national security concerns with the economic realities faced by U.S. tech giants.