Trump Invokes National Security Law to Levy 25% Tariff on Advanced AI Chips, With Broad U.S. Exemptions

President Donald Trump has opened a new front in his trade and technology agenda, invoking national security powers to impose a 25% tariff on some of the world’s most advanced computer chips while largely shielding U.S. data centers and artificial intelligence startups from the cost.

In a proclamation signed Jan. 14 and taking effect at 12:01 a.m. Eastern time Jan. 15, Trump ordered a 25% ad valorem duty on a narrow class of high-end “advanced computing chips” and certain products that contain them. The measure is grounded in Section 232 of the Trade Expansion Act of 1962, the same national security statute he used in his first term to tax steel and aluminum imports.

The move marks the first time a U.S. president has used Section 232 to target the sophisticated semiconductors that power AI models and data centers, rather than bulk industrial commodities. It is designed to make foreign-bound shipments of Nvidia and Advanced Micro Devices’ latest datacenter-class chips more expensive at the U.S. border, even as the administration carves out broad exemptions for hardware used inside the United States.

Tariff framed as security measure

The proclamation, titled “Adjusting Imports of Semiconductors, Semiconductor Manufacturing Equipment, and Their Derivative Products Into the United States,” follows a Section 232 investigation by the Department of Commerce. In a report delivered Dec. 22, 2025, Commerce concluded that imports of semiconductors, chip-making equipment and related products “are being imported into the United States in such quantities and under such circumstances as to threaten to impair the national security of the United States.”

Trump adopted that finding in the proclamation, declaring semiconductors “essential to the United States’ economic, industrial, and military strength.” He noted that chips underpin all 16 critical infrastructure sectors identified in a 2024 national security memorandum, from energy and transportation to communications and health care.

The document cites a long-running concern in Washington: the gap between U.S. leadership in chip design and its dependence on foreign factories. The United States accounts for roughly a quarter of global semiconductor consumption but “fully manufactures only approximately 10 percent of the chips it requires,” the proclamation states.

Commerce Secretary Howard Lutnick recommended a two-phase response. The first phase, now in force, imposes what the proclamation calls a “very narrow” 25% tariff on certain advanced computing chips and their derivatives, while U.S. officials pursue negotiations with major semiconductor-producing countries. The second phase, left for the future, contemplates “significant” and broader tariffs on a wider array of chips and potentially on semiconductor manufacturing equipment, coupled with incentives for companies that build production capacity in the United States.

Aimed at AI-class hardware

The White House did not name specific manufacturers in the legal text, but a fact sheet released alongside the proclamation pointed to Nvidia’s H200 graphics processing unit and AMD’s MI325X accelerator as examples of the targeted products. Those chips sit at the core of large AI training clusters and are produced predominantly by Taiwan Semiconductor Manufacturing Co. in Taiwan.

According to reporting by Reuters and other outlets, certain systems and devices that integrate those chips and meet performance thresholds may also fall under the new duty. The 25% tariff applies to covered goods “entered for consumption, or withdrawn from warehouse for consumption” in the United States on or after the effective time, and importers cannot receive refunds of the duty through the usual drawback mechanisms.

The measure is not country-specific on its face, meaning it applies to covered products regardless of where they are made. In practice, it is expected to fall most heavily on chips fabricated in East Asia and imported into the United States either for use in domestic data centers or for re-export to foreign customers, including in China.

Broad exemptions for U.S. users

While the headline rate is high, the proclamation is studded with exceptions. The 25% tariff does not apply when covered chips are imported for use in U.S. data centers; for research and development activities in the United States; for repairs or replacements performed domestically; or for use by U.S. startups.

Additional carve-outs cover non-data-center consumer uses such as high-end personal computers and workstations, non-data-center civil industrial applications and U.S. public-sector deployments. The Commerce secretary is also authorized to grant further exemptions where imports are deemed to support the domestic technology supply chain or U.S. manufacturing capacity for derivative products.

Those carve-outs mean major American cloud providers and AI companies building infrastructure inside the country are largely insulated from the new duty, at least for now. The tariff will instead bite hardest on flows of high-end AI chips that cross the U.S. customs boundary on their way to foreign buyers.

Intersection with China policy

The proclamation lands alongside a separate policy shift affecting Nvidia’s business with China. In recent months, the Trump administration has approved exports of the H200 to Chinese customers under strict conditions, following earlier export controls that restricted even slightly downgraded versions of Nvidia’s datacenter GPUs.

Under the new arrangement, described by people familiar with the matter and first reported by financial media, U.S. authorities allow some shipments of H200-class chips to China, subject to third-party testing, know-your-customer requirements and limits on the volume sold relative to U.S. domestic supply. The Section 232 tariff now ensures that when those chips are routed through the United States, customs officials collect a 25% duty at the border.

Supporters inside the administration argue that structure preserves U.S. visibility into where advanced chips end up, channels revenue toward U.S. coffers and maintains leverage over future exports. Critics in the national security community have questioned whether conditioning and taxing those flows is sufficient to offset the strategic boost they may give to Chinese AI capabilities.

Pressure and reassurance for allies

Major semiconductor-producing allies are watching the new measure closely. South Korea’s industry minister, Kim Jung-kwan, said in Seoul on Jan. 15 that his government would “keep monitoring” the impact of U.S. tariffs on AI chips and aim to minimize any damage to Korean firms such as Samsung Electronics and SK Hynix. After consulting with chipmakers, the ministry assessed the immediate impact as limited because hardware destined for U.S. data centers and startups is exempt, but it warned of “significant uncertainty” if Washington moves ahead with broader tariffs on a wider range of semiconductors.

For Taiwan, the crackdown is intertwined with a substantial industrial and trade negotiation. TSMC is already building multiple fabrication plants in Arizona backed by grants and loans under the CHIPS and Science Act of 2022, which set aside $52.7 billion to bolster U.S. semiconductor manufacturing and research. U.S. and Taiwanese officials are also working on a trade arrangement that, according to people briefed on the talks, would reduce U.S. tariff rates on Taiwanese goods in exchange for hundreds of billions of dollars in additional investment by Taiwanese firms, potentially including a dozen or more fabs on U.S. soil.

The prospect of a second-phase Section 232 action imposing sweeping tariffs on all imported chips and some manufacturing tools has become a key piece of leverage in those discussions, according to trade analysts. Companies that commit to making advanced chips in the United States are expected to receive favorable treatment or exemptions if that broader action goes forward.

Industrial policy by tariff

Lutnick, a former Wall Street executive who took office as commerce secretary in February 2025, has been explicit about using tariffs as a central tool of economic policy. In past remarks, he has said tariffs are “worth it” even if they risk slowing growth, arguing that they are necessary to restore manufacturing jobs in strategic industries.

By combining Section 232 tariffs with CHIPS Act subsidies and export controls, his department has effectively turned Commerce into the hub of a more assertive industrial strategy. The new proclamation instructs Commerce, the U.S. International Trade Commission and U.S. Customs and Border Protection to define the specific tariff codes that will capture the targeted AI-class chips and to set up end-use certification systems to administer the exemptions. Products entering U.S. foreign trade zones will have to be admitted in “privileged foreign status,” ensuring the 25% duty is collected if they enter the U.S. market.

The proclamation also orders Commerce to report back by July 1 on the market for semiconductors used in U.S. data centers, a review that could support either narrowing or broadening the tariff’s scope.

A test case for the next phase

For now, most consumer electronics and many lower-end chips are untouched. But the administration has repeatedly signaled that this is only the first step. Trump floated tariffs of up to 100% on some semiconductor imports during the 2024 campaign and early in his second term, and the Section 232 investigation that underpins the current move covers not only AI accelerators but also a wide range of chips and chip-making equipment.

Business groups and foreign governments are bracing for the possibility that a much wider swath of the $600 billion global semiconductor trade could be drawn into the tariff net. Economists say such a step could push up prices across industries that rely on chips, from autos and factory machinery to medical devices and consumer electronics, and strain relations with key allies.

Trump’s latest proclamation stops short of that broad action. Instead, it targets a thin but strategically important slice of the market—the AI-class processors that have become the most coveted components in technology—while giving Washington new tools to steer where those chips are made and who can afford to use them.

In doing so, it signals that the silicon driving advances in artificial intelligence will increasingly move across borders not only as commercial goods, but as instruments of national strategy, subject to security reviews, negotiated exemptions and, now, a sizable toll at the U.S. border.