US Implements New Export Controls on Remote AI Chip Access via Third-Country Data Centers
The US Department of Commerce issued new regulations on November 16, 2026, to prevent third-country remote access to advanced AI chips. These rules aim to stop Chinese companies from bypassing existing export controls through remote computing arrangements located in third countries. The Department of Commerce stated that the Department of the Treasury and the National Security Council drafted and directed the new measures.
New licensing requirements for data centers
Under the new regulations, data center operators in Singapore, Malaysia, Thailand, Indonesia, and Brazil must obtain US export licenses to purchase advanced AI chips. The regulations specifically target data centers with a Total Processing Performance (TPP) exceeding 12 million. This performance threshold is roughly equivalent to the capacity of 750 Nvidia H100 or H200 chips. To prevent evasion of controls, the new measures require exporters and global data center operators to conduct enhanced due diligence regarding the ownership structures of their customers.
The regulations also extend requirements to semiconductor manufacturing. Chinese chip design companies must now obtain US export licenses when manufacturing chips using 16/14nm or advanced process nodes at TSMC or Samsung. These measures are intended to tighten the oversight of high-end computing power availability.
Impact on regional investments and financing
The new regulations have already impacted large-scale technology financing. A 4.8 billion USD loan involving Aolani Cloud Pte has reportedly stalled. This loan was intended to fund the purchase of Nvidia GPUs for a Malaysian data center designed to serve ByteDance. Five unnamed banks that were evaluating the Aolani Cloud loan are reportedly discussing non-participation or internal exits due to compliance risks associated with the new rules.
An Aolani spokesperson said the company is working with regulators and financing partners to ensure project progress complies with new regulations. However, the regulatory shift introduces significant uncertainty for regional infrastructure projects. Economist Robert Walker of the Lowy Institute said the new restrictions will pose significant resistance to future data center investments in the affected countries.
Global computing and revenue implications
The shift in regulation comes as the demand for high-end AI computing power remains high. Experts suggest that 30% of China's high-end AI computing power is expected to come from overseas data center rentals by 2027. This trend highlights the importance of the new US controls in managing how advanced hardware is utilized globally.
The economic implications for hardware manufacturers remain a subject of debate. While some estimates suggest a potential annual revenue risk for Nvidia between 18 billion and 35 billion USD, these figures are subject to different estimates. The regulatory environment continues to evolve as the US government implements stricter controls to prevent Chinese entities from accessing advanced AI computing power through third-party data centers in Southeast Asia and Brazil.
Glossary
- Nvidia H100 — The Nvidia H100 is a chip used in computing. Its capacity is roughly equivalent to 750 chips when compared to a performance threshold of 12 million Total Processing Performance.
- advanced process nodes — Advanced process nodes are used in semiconductor manufacturing. They are targeted by specific regulations regarding chip design and manufacturing.
- TSMC — TSMC is a semiconductor manufacturing company. It is one of the locations where chip design companies may manufacture chips.
- Samsung — Samsung is a technology company. It is a location where chip design companies may manufacture chips.
This article was written with AI assistance from multiple news reports and is reviewed by our editors under our editorial principles. Editorial Standards · Tips & Contact
Comments 0