Exporting U.S.-made artificial intelligence (AI) chips is subject to strict regulations enforced by the U.S. Department of Commerce’s Bureau of Industry and Security (BIS) under the Export Administration Regulations (EAR). These restrictions primarily aim to protect national security by limiting adversaries’ access to advanced AI technologies while balancing U.S. economic interests.

The U.S. divides the world into three tiers for AI chip export controls, each with different licensing requirements.
Tier 1: Close Allies (Unrestricted Access) Exports of advanced AI chips to 18 allied countries are generally unrestricted, provided the recipient entity is headquartered in one of these countries and not owned by a parent company in a non-Tier 1 country. These countries include: Australia, Belgium, Canada, Denmark, Finland, France, Germany, Ireland, Italy, Japan, Netherlands, New Zealand, Norway, South Korea, Spain, Sweden, Taiwan, and the United Kingdom. A new License Exception Artificial Intelligence Authorization (AIA) allows exports, re-exports, or in-country transfers to these countries without a license.
Under Tier 2: Middle-Tier Countries (Capped Access) Approximately 150 countries, including India, Israel, Singapore, and the United Arab Emirates, face caps on the number of AI chips they can import. A cumulative maximum of 790 million Total Processing Performance (TPP) is allowed for these countries through 2027, equivalent to about 50,000 Nvidia H100 GPUs. Exports require a license, reviewed with a presumption of approval until the country’s cap is reached, after which applications face a presumption of denial. Companies can apply for National Verified End User (NVEU) status, allowing them to receive chips up to a per-company, per-country quarterly cap (max 25% of their total AI computing power to middle-tier countries). Small-scale orders (up to ~1,700 GPUs) for universities, medical institutions, or research organizations are exempt from caps via the License Exception Low Processing Performance (LPP).

Under Tier 3: Arms-Embargoed Countries (Prohibited Access) Exports of advanced AI chips to approximately 24 countries subject to U.S. arms embargoes (Country Group D:5) and Macau are strictly prohibited, with license applications reviewed under a presumption of denial. These countries include: China, Russia, Iran, North Korea, Cuba, Belarus, Syria, Venezuela, and others. This catagory aims to block adversaries from accessing U.S. AI chips directly or through third countries.

The restrictions apply to advanced AI chips, specifically graphics processing units (GPUs) and integrated circuits (ICs) classified under Export Control Classification Numbers
The regulations extend to foreign-produced items that are Derived from U.S.-origin software, technology, or equipment (e.g., ICs or servers). Subject to the EAR if produced by a plant or major component under specified ECCNs. This AI Model Weights FDP Rule ensures that even non-U.S.-made AI model weights or chips using U.S. technology are restricted when exported to non-Tier 1 countries.

Exporters must conduct due diligence to ensure chips are not diverted to prohibited countries or end-users. The Foundry Due Diligence Rule (January 2025) provides guidance on compliance procedures.

Companies like Nvidia and Oracle have criticized the rules, arguing they threaten U.S. competitiveness by limiting markets for U.S. firms while allowing foreign competitors (e.g., China’s Biren Technology) to gain market share.

Countries like Israel and Poland have raised concerns, with Israel noting strategic repercussions for its tech sector (20% of GDP). The EU has objected to restrictions on members like Portugal, arguing they pose no security risk.

The January 2025 rules build on earlier controls from October 2022 and 2023, which targeted China’s access to advanced semiconductors. These were expanded in December 2024 to include 140 companies on the Entity List and restrict high-bandwidth memory. The Biden administration’s “small yard, high fence” strategy focuses on controlling key chokepoints in the AI supply chain. However, critics argue the broad scope of the 2025 rules risks economic harm without fully curbing adversaries’ capabilities. The incoming Trump administration (as of January 2025) may revise or repeal these rules, given its prior support for AI development and criticism from industry leaders.
The rules aim to maintain a U.S. lead in AI (estimated at 6-18 months over China), but their effectiveness depends on multilateral cooperation.

Anthropic, a leading AI company has supported the stand. Company in its blogpost wrote, “The development of AI systems with capabilities rivaling Nobel Prize winners across multiple disciplines will transform our economy, national security, and society. By strengthening—not weakening—the Diffusion Framework and vigorously enforcing export controls, America can ensure that these transformative technologies are developed domestically, by American firms, and in alignment with American values and interests. Our continued leadership in AI depends on maintaining and expanding our compute advantage through decisive policy action today.”

Galactik Views