US eyes broad new restrictions on chip exports - AI News Today Recency

# US Eyes Broad New Restrictions on Chip Exports: What's Next for the Global AI Supply Chain

The United States is reportedly considering sweeping new chip export controls that would require government approval for shipping advanced AI semiconductors anywhere outside the country[7]. This potential regulatory shift marks another significant escalation in America's efforts to maintain technological dominance while protecting national security interests in the rapidly evolving AI sector.

Trump Administration Shifts Strategy on Advanced Semiconductor Controls

The Trump administration has already implemented substantial changes to semiconductor policy in recent weeks. On January 14, 2026, the administration fundamentally shifted its approach by allowing exports of advanced AI chips to China under specific conditions, moving from a "presumption of denial" to a "case-by-case review" policy for NVIDIA H200 and AMD MI325X-equivalent chips[4]. However, this apparent liberalization came with significant caveats: the administration simultaneously imposed a 25 percent tariff on advanced AI chip imports that are not destined for the US technology supply chain[2][5].

The rationale behind these measures centers on national security. The White House determined that the United States' capacity to produce semiconductors and advanced semiconductor manufacturing equipment is insufficient to meet domestic demand, creating dangerous reliance on foreign sources[2]. This vulnerability has prompted the administration to pursue a dual strategy: allowing certain exports while simultaneously protecting domestic manufacturing capacity through tariffs and import restrictions.

Congressional Opposition and the AI Overwatch Act

The policy shift has sparked significant congressional pushback. Representative Molenaar, chair of the House Select China committee, raised concerns about the practical feasibility of implementing new export licensing requirements, particularly given current global shortages of critical components[3]. More dramatically, Congress passed the "AI Overwatch Act" on January 22, 2026, which introduces a permanent state of uncertainty for the global supply chain by allowing the legislature to revoke licenses granted by the Commerce Department at any time[4].

This legislative action reflects deep concerns that easing restrictions on advanced chip exports to China could erase America's decisive AI advantage. Chinese firms have already placed orders for over 2 million H200 chips, valued at up to $14 billion, which could enable China to significantly close the technological gap with US AI leading labs[4].

China's Mineral Leverage and Geopolitical Calculations

The policy shift also reflects broader geopolitical pressures. China controls approximately 70 percent of the world's refined silver used in chips and dominates the processing of critical rare earth elements such as gallium, essential materials for advancing chip production[4]. Since July 2023, China has progressively restricted exports of these critical minerals, escalating to outright export bans to the United States in December 2024. This mineral leverage has created significant pressure on the US to soften its stance on chip exports, demonstrating how semiconductor policy has become intertwined with broader supply chain vulnerabilities.

Tariff Structure and Implementation Challenges

The administration's regulatory framework creates a complex, layered tariff structure that extends beyond the immediate AI chip sector. The 25 percent tariff on advanced computing chips applies specifically to imports that do not contribute to building out the US technology supply chain[2]. However, the broader semiconductor tariff landscape includes legacy Section 301 tariffs on Chinese-origin semiconductors, reciprocal tariffs, and heightened reciprocal tariffs suspended through November 10, 2026[6].

Critical deadlines loom for companies navigating this environment. The suspension of heightened reciprocal tariffs expires on November 10, 2026, unless extended, and a new semiconductor tariff rate is scheduled to increase on June 23, 2027[6]. Companies importing advanced AI chips before selling to foreign customers face the 25 percent tariff even if those chips are destined for allied nations, creating additional compliance complexity[5].

Future Negotiations and International Agreements

The administration has signaled openness to negotiated solutions. A January 15, 2026 agreement with Taiwan provides a potential model: Taiwanese semiconductor companies investing in US production may import up to 2.5 times their planned capacity without paying Section 232 tariffs, or 1.5 times their new US production capacity for completed projects[3]. The administration indicated it may consider similar arrangements with other countries, suggesting that negotiations could reshape the tariff landscape significantly.

Additionally, the Commerce Department is required to provide the President with an update on the semiconductor market used in US data centers by July 1, 2026, potentially triggering further policy adjustments[2]. Within 90 days of the proclamation's issuance, the Secretary and Trade Representative must also update the President on the status of ongoing negotiations regarding semiconductor imports[2].

Frequently Asked Questions

What specific chips are subject to the new 25 percent tariff?

The 25 percent tariff applies to advanced AI chips covered by the President's January 14, 2026 proclamation, specifically advanced computing chips that are not destined for the US technology supply chain[2][5]. NVIDIA H200 and AMD MI325X-equivalent chips are the primary focus of the new licensing and tariff policies[4].

Can companies export AI chips to China under the new rules?

Yes, but only under strict conditions. The Trump administration shifted to a "case-by-case review" policy for NVIDIA H200 and AMD MI325X chips exported to China, replacing the previous "presumption of denial" approach[4]. However, exporters must meet onerous certification requirements and provide extensive documentation regarding US and global supply chains and Chinese end-user information[3].

What is the AI Overwatch Act, and how does it affect chip exports?

The AI Overwatch Act, passed on January 22, 2026, grants Congress the power to revoke export licenses granted by the Commerce Department at any time[4]. This creates permanent uncertainty for companies relying on export approvals, as legislative action could suddenly reverse granted licenses.

When will the new semiconductor tariffs take effect, and what are the key deadlines?

The 25 percent tariff on advanced AI chip imports took effect on January 15, 2026[5]. Two critical upcoming deadlines are November 10, 2026, when heightened reciprocal tariffs suspension expires unless extended, and June 23, 2027, when a new semiconductor tariff rate is scheduled to increase[6].

How does the Taiwan agreement serve as a model for other countries?

Under the January 15, 2026 agreement with Taiwan, Taiwanese semiconductor companies investing in US production can import up to 2.5 times their planned capacity without paying Section 232 tariffs, or 1.5 times their new US production capacity for completed projects[3]. The administration indicated similar arrangements may be negotiated with other countries to incentivize domestic manufacturing investment.

Why is the US restricting chip exports if it also wants to allow sales to China?

The administration is pursuing a dual strategy to address competing priorities: maintaining technological dominance through selective export controls while also leveraging chip sales as a negotiating tool with China[5]. The restrictions on imports not destined for the US supply chain aim to protect domestic manufacturing capacity, while case-by-case review of China exports allows the administration flexibility in geopolitical negotiations[2][4].