
US Tightens AI Chip Export Rules: 12M TPP Cap and Global Data Center Licenses
By Ali Al-Montaser — Senior AI Technical Editor
The Trump administration issued an interim final rule requiring Commerce Department licenses for AI chip and server exports to any global data center that may serve Chinese, Russian, Iranian, North Korean, or 18 other restricted countries. The rule applies to any network exceeding 12 million TPP—roughly 750 Nvidia H100 or H200 chips—and adds new licenses for Malaysia, Singapore, Indonesia, Thailand, and Brazil, with a 90-day compliance window for existing operators. It repeals the Biden-era AI Diffusion Rule and targets indirect compute access via brokers and intermediaries.
Executive Overview
The Trump administration on Monday announced an interim final rule that requires the US Commerce Department to grant licenses for the export, reexport, or transfer of AI chips and servers to any data center worldwide that may provide computing capabilities to companies in China or other restricted countries. The rule takes effect immediately upon publication, with a 90-day compliance window for existing data center operators. It expands export controls beyond direct chip shipments to prevent remote access to AI infrastructure, targeting firms in the restricted D:5 group—China, Russia, Iran, North Korea, and 18 other countries. The move follows reports of Chinese firms seeking external compute and repeals the Biden-era AI Diffusion Rule.
📊 Official Technical Specifications & Data Sheet
| Technical Aspect | Confirmed Official Data |
|---|---|
| 💰 Pricing & Usage Cost | The rule does not include direct pricing; operational cost consists of export license fees and data center operator compliance. No fees are imposed on end users per se. |
| 🌐 Platforms & Immediate Availability | Applies globally to any data center or network providing compute exceeding 12 million TPP. Additional licenses required for Malaysia, Singapore, Indonesia, Thailand, and Brazil. Does not apply to chips destined for US-headquartered companies. |
| ⚡ Performance & Speed Benchmarks | The 12 million TPP threshold is roughly equivalent to the maximum power of about 750 Nvidia H100 or H200 chips. The rule aims to prevent training and operation of advanced AI models. |
| 🛡️ Security & Breach Resistance | Requires exporters and data center operators to monitor customer ownership structures and activities to ensure they are not managed by any entity in the D:5 group. Covers indirect access via intermediaries or brokers. |
| 🧠 Context Window | Not applicable — the rule regulates hardware exports, not language model specifications. |
| 🌍 Arabic Language & Regional Support | The rule contains no Arabic-specific provisions. It applies to data centers in Gulf and Middle East countries if they provide compute to any D:5 company or exceed the 12 million TPP threshold. |
Deep-Dive Features & Architecture
The interim rule significantly broadens the scope of US export controls from merely shipping chips to Chinese companies to preventing remote access to AI infrastructure. Under the text, the Commerce Department must grant licenses for the export, reexport, or transfer of AI chips and servers to data centers anywhere in the world if they are likely to provide compute to companies in China or other restricted countries. The rule targets firms in the restricted D:5 group, which includes China, Russia, Iran, North Korea, and 18 other countries.
The rule sets a threshold of 12 million TPP, an export control metric used by the US to measure computing capacity, roughly equivalent to the maximum power of about 750 Nvidia H100 or H200 chips. It also covers indirect access through intermediaries or brokers and requires exporters and data center operators to monitor customer ownership structures and activities to ensure they are not managed by any entity in the D:5 group. Additionally, the rule adds new licensing requirements for the export, reexport, or transfer of advanced semiconductors to Malaysia, Singapore, Indonesia, Thailand, and Brazil, because the Commerce Department said they "represent higher-than-usual diversion risks," with an exception for chips "destined" for US-headquartered companies.
The move comes days after the resignation of Jensen Huang, Nvidia's co-founder, amid a scandal involving an alleged Chinese spy linked to a client of the chipmaker. Reports indicated that the US government wanted to add that client, Megaspeed, to a special entity list because it believed the company was buying thousands of Nvidia chips to provide compute to Chinese firms in circumvention of existing export restrictions. Reuters also reported last week that Chinese AI companies are racing to secure computing capacity abroad amid fears of stricter US restrictions.
Benchmark & Competitive Performance
The interim rule repeals the AI Diffusion Rule issued by the Biden administration in January 2025, which sought to impose country-level limits on advanced AI chip exports and restrict the transfer of certain powerful AI models. In May, the Trump administration said it would discard that framework, arguing it added unnecessary restrictions on US companies and allies. By comparison, the new rule replaces country-level caps with a compute-threshold approach (12 million TPP) and extends controls to remote access and intermediary channels, aiming to close loopholes that allowed Chinese firms to access US AI infrastructure indirectly.
Industry Impact & Enterprise Adoption
The rule is expected to significantly impact global data center operators, cloud providers, and AI companies that serve multinational clients. Enterprises with data centers in the Middle East, Southeast Asia, or other regions may need to obtain licenses or restructure their operations to comply. The 90-day compliance window gives existing operators time to adapt, but the broad scope—covering any network exceeding 12 million TPP—could affect a wide range of commercial and research facilities. The rule does not apply to commercial recommendation systems used in e-commerce or social media, according to the published text, but its implications for AI training and inference workloads are substantial.
Conclusion
The Trump administration's interim final rule marks a significant escalation in US efforts to control the global spread of advanced AI computing, shifting from chip-level export restrictions to a comprehensive licensing regime for data centers worldwide. With a 12 million TPP threshold, new licenses for five additional countries, and a 90-day compliance deadline, the rule aims to prevent Chinese and other restricted entities from accessing remote compute. As the AI chip landscape evolves, this regulatory move will likely reshape global AI infrastructure deployment and compliance strategies for years to come.
Media Source: TechCrunch AI | Fact Verification & Analysis: AI Tools Oasis
Frequently Asked Questions
12 million TPP is a performance metric used by the US to measure computing capacity, roughly equivalent to the maximum power of about 750 Nvidia H100 or H200 chips. The rule applies to any data center or network that can provide more than 12 million TPP.
The rule targets companies in the restricted D:5 group, which includes China, Russia, Iran, North Korea, and 18 other countries. It also adds new licensing requirements for advanced chip exports to Malaysia, Singapore, Indonesia, Thailand, and Brazil due to higher-than-usual diversion risks.
Existing data center operators whose services fall under the new rule must obtain licenses within 90 days of the rule's effective date.
Yes, the interim final rule repeals the AI Diffusion Rule issued by the Biden administration in January 2025, which imposed country-level limits on advanced AI chip exports and restricted the transfer of certain powerful AI models.
No, the rule does not apply to commercial recommendation systems such as those used in e-commerce or social media, according to the published text.

AI Tools Oasis Team
Bringing you the latest news and analysis in the world of Artificial Intelligence with accuracy and credibility. Follow us for all updates.


