Trump Imposes 25% Tariff on AI Chip Imports

• Trump tariffs target advanced AI processors, including Nvidia H200 and AMD MI325X.
• Policy creates contradiction: H200 exports to China were approved on Tuesday, subjected to 25% import duties on Wednesday.

President Trump has imposed a 25% tariff on imports of artificial intelligence chips, including Nvidia’s H200 processor and AMD’s MI325X, just one day after his administration approved H200 exports to China – creating a policy whiplash that reflects competing priorities between boosting domestic semiconductor manufacturing and maintaining American chipmakers’ global competitiveness.

The tariff, announced on Wednesday through a White House national security proclamation, follows a nine-month investigation under Section 232 of the Trade Expansion Act and targets AI semiconductors meeting specific performance benchmarks. The move aims to incentivise onshore chip production at a time when the US manufactures only approximately 10% of the semiconductors it requires, according to the White House.

The timing creates immediate complications for Nvidia’s China strategy. Under export rules announced Tuesday by the Bureau of Industry and Security (BIS), all H200 chips destined for China must transit through the US from Taiwan for mandatory third-party laboratory testing.

When these chips enter US territory for verification, they now trigger the 25% tariff before reaching Chinese customers – effectively adding high cost to exports that the administration just approved.

Tariff scope and exemptions for AI chip imports

The Section 232 tariff applies to advanced semiconductors meeting specific technical thresholds for AI processing abilities, as well as devices containing these chips. However, the White House has carved out exemptions designed to avoid disrupting domestic technology infrastructure.

According to the White House fact sheetthe 25% tariff will not apply to AI chips imported for US data centres, startups, non-data centre consumer applications, non-data centre civil industrial applications, or US public sector use. Commerce Secretary Howard Lutnick retains discretion to grant additional exemptions on a case-by-case basis.

The exemption structure suggests the administration’s primary target is commercial imports that could be manufactured domestically, not chips supporting US-based AI development and deployment.

However, chips routed through the US for export to foreign markets – like the China-bound H200 processors requiring mandatory testing – appear to fall in the tariff’s scope.

The proclamation clarifies that the 25% AI chip tariff would not stack on top of other Trump administration tariffs imposed under separate Section 232 orders. Affected semiconductors would be exempt from duties on copper, aluminium, steel, and auto and truck parts.

Market reaction and industry silence

Nvidia, AMD, and Qualcomm shares all traded slightly lower in after-hours trading following the tariff announcement.Nvidia did not respond to requests for comment on how the company plans to navigate the conflicting policy signals – export approval followed immediately by tariff imposition.

AMD issued a brief statement confirming the company “complies with all US export control laws and policies,” but did not address the tariff’s business implications. Taiwan Semiconductor Manufacturing Company, which manufactures chips for Nvidia, AMD, and other US designers, also did not respond to comment requests.

Broader semiconductor tariff strategy and domestic production goals

The AI chip tariff represents one component of a broader Trump administration push to re-shore semiconductor manufacturing. The White House fact sheet indicates that Trump may impose additional tariffs on semiconductor imports and derivative products in the near future to further encourage domestic production.

The administration’s proclamation emphasises that current US reliance on foreign chip manufacturing – particularly Taiwan Semiconductor Manufacturing Company’s dominance in advanced node production – constitutes a “economic and national security risk.”

The Section 232 investigation concluded that incentivising domestic semiconductor production serves essential national security objectives. However, the tariff announcement comes amid separate policy developments that complicate the domestic production narrative.

Trump in December announced plans to impose tariffs on Chinese semiconductor imports over Beijing’s “unreasonable” pursuit of chip industry dominance, but delayed that action until June 2027 following a year-long Section 301 unfair trade practices investigation launched by the Biden administration.

The December announcement focused on “legacy” or older-technology chips, while Wednesday’s tariff targets cutting-edge AI processors – suggesting the administration is pursuing a multi-pronged approach to semiconductor trade policy.

Constitutional questions and export tax concerns

Trump stated in December that he would allow Nvidia to sell H200 chips to China in exchange for a cut of sales revenue – a proposal that legal experts questioned as potentially violating the US Constitution’s ban on taxing exports.

The 25% tariff announced on Wednesday may represent an alternative mechanism to achieve similar revenue objectives through import duties not export taxes. However, the White House made no mention of revenue sharing in the official proclamation, and the tariff applies to AI chip imports not specifically targeting China-bound exports.

The routing requirement for China-bound chips – manufactured in Taiwan, entering the US for testing, then exported to China – creates a unique situation where the same chip is subject to import tariffs despite ultimately being exported. The structure allows the administration to impose costs on China-destined semiconductors without directly taxing exports.

Implications for Asia-Pacific technology markets

For technology companies and data centre operators in the Asia-Pacific region, the tariff announcement compounds existing uncertainty around AI chip access and pricing.

Chinese technology firms already face scepticism from their own government about purchasing Nvidia processors, with Beijing indicating it will approve H200 orders only for university research and development laboratories not commercial deployment.

The 25% tariff adds a high cost to China-bound H200 chips that must transit the US for mandatory testing, potentially pricing them out of reach even for the limited university use cases Beijing has indicated it might approve. This could accelerate Chinese technology companies’ shift toward domestic semiconductor alternatives from manufacturers like Huawei, Moore Threads, and other local chipmakers.

Broader APAC markets not subject to the US-China export restrictions may see AI chip costs rise if manufacturers pass through tariff expenses, though the exemption for data centre imports suggests most regional cloud providers and technology infrastructure operators should avoid direct tariff impact.

The policy development underscores the volatility facing technology supply chains operating between the US and Asian markets. Companies planning AI infrastructure investments must now account for rapidly shifting trade policies that can fundamentally alter cost structures and chip availability in 24-hour periods.

The tariff takes effect immediately, while the H200 export approval policy announced Tuesday takes effect January 16, 2026.

Want to experience the full spectrum of enterprise technology innovation? Join TechEx in Amsterdam, California, and London. Covering AI, Big Data, Cyber Security, IoT, Digital Transformation, Intelligent Automation, Edge Computing, and Data Centres, TechEx brings together global leaders to share real-world use cases and in-depth insights. Click here for more information.