WASHINGTON, Nov 16: The United States has announced tighter restrictions on access to advanced artificial intelligence chips, introducing new export control requirements aimed at preventing Chinese companies from obtaining high-performance computing capabilities through overseas data centres.
The move, announced by the US Commerce Department on Monday, seeks to close regulatory gaps that have allowed companies based in China to access advanced semiconductor technology through computing facilities operating in other countries. The restrictions come amid increasing concerns in Washington over the potential use of sophisticated AI processors in military applications, advanced surveillance and other sensitive technologies.
Under the new regulations, shipments of certain advanced semiconductor chips to data centres located in several countries will require approval through US export licences. The policy also introduces additional compliance obligations for overseas facilities providing substantial computing capacity to businesses headquartered in China.
The countries identified as higher-risk destinations for potential diversion include Singapore, Thailand, Malaysia, Indonesia and Brazil. Shipments of covered advanced processors to data centres in these locations will be subject to licensing requirements, although the provisions do not apply to US companies operating their own facilities in those countries.
The administration’s decision reflects a broader effort to prevent restricted technology from reaching Chinese artificial intelligence developers through indirect commercial arrangements. Instead of focusing exclusively on direct exports to Chinese buyers, the revised framework also addresses the international infrastructure through which advanced computing resources can be accessed remotely.
Modern AI systems require enormous computing capacity to process information, train models and perform increasingly complex tasks. High-performance graphics processing units and specialised AI accelerators are particularly important because they can handle large volumes of calculations simultaneously.
These processors are used by technology companies to develop generative AI applications, train large language models, operate automated systems and support advanced research. Their importance has increased as businesses and governments invest heavily in artificial intelligence across industries.
US officials have argued that unrestricted access to the most advanced chips could create national security concerns, particularly when computing infrastructure is used by companies operating in countries considered strategic competitors.
The new measures also require certain overseas data centres supplying significant amounts of computing power to Chinese-headquartered businesses to obtain export authorisations within 90 days. Facilities failing to meet the requirements could face additional restrictions on their ability to access US-origin technology.
However, the rules contain exemptions for systems used by online retail and social media companies, which the Commerce Department said do not present the same level of national security concern under the framework.
The latest announcement comes as Washington increases scrutiny of the international semiconductor supply chain and the movement of advanced processors through third-party markets.
One of the major challenges facing US regulators has been determining how to monitor the use of chips after they have been shipped outside American territory. Data centres can provide computing services to customers located thousands of kilometres away, making it difficult to establish who ultimately benefits from the hardware.
The growing popularity of cloud computing has further complicated the issue. Companies no longer need to purchase expensive processors or establish their own infrastructure to develop sophisticated AI systems. Instead, they can rent computing capacity from specialised service providers and access powerful hardware through internet-based platforms.
This arrangement has created new commercial opportunities but has also raised questions about the effectiveness of traditional export-control systems.
The revised policy attempts to address this challenge by extending regulatory oversight to certain overseas computing facilities and service arrangements. It is intended to make it more difficult for Chinese companies to bypass restrictions by obtaining computing resources through businesses operating in other markets.
The announcement also formally withdraws a Biden-era regulation that had sought to limit worldwide access to advanced AI chips through a broader licensing framework. The earlier rule was not enforced and had been expected to be replaced by a different regulatory approach.
The timing of the decision is particularly significant as US President Donald Trump prepares for discussions with Chinese President Xi Jinping in Shenzhen during the Asia-Pacific Economic Cooperation summit.
Technology and trade relations between Washington and Beijing remain major areas of disagreement. Semiconductor restrictions, access to advanced manufacturing equipment, artificial intelligence development and the security of critical supply chains have become central issues in the wider economic relationship.
The latest announcement also follows controversy surrounding the alleged diversion of advanced processors through overseas technology companies. A Singapore-based cloud computing firm, Megaspeed, has faced scrutiny over concerns that restricted Nvidia chips may have been diverted to Chinese users.
The company has maintained that it complies with applicable laws and export-control requirements.
The controversy has increased pressure on US authorities to demonstrate that existing safeguards are effective and that companies cannot easily circumvent restrictions by moving computing operations through third countries.
The new policy is expected to place additional responsibilities on semiconductor suppliers, cloud service providers and operators of large-scale data centres. Businesses may need to strengthen customer verification procedures, review existing contracts and establish clearer systems for monitoring how computing resources are used.
For technology companies operating internationally, the rules could introduce additional compliance costs and affect decisions regarding the location of future infrastructure investments.
Data centre operators may also have to examine the ownership structures and business relationships of customers seeking access to advanced processors. Companies providing computing services to multiple international clients could face greater scrutiny over whether their facilities are being used to support restricted activities.
The semiconductor industry is likely to monitor the implementation of the regulations closely, particularly because advanced chips are among the most important components of the global AI economy.
Companies such as Nvidia have played a central role in supplying processors used for artificial intelligence training and inference. Their products are widely deployed by cloud providers, research institutions and technology developers.
The new restrictions could influence how manufacturers distribute their products and how international customers arrange access to computing infrastructure. Businesses may increasingly look for alternative suppliers, establish facilities in different locations or develop more efficient computing systems that require fewer advanced processors.
For China, the restrictions represent another challenge in its efforts to strengthen domestic semiconductor manufacturing and reduce dependence on foreign technology.
Chinese technology companies have invested heavily in developing homegrown processors, AI models and computing infrastructure. However, producing advanced chips at competitive performance levels remains a complex industrial undertaking involving specialised equipment, materials, manufacturing expertise and substantial capital investment.
The latest US measures could encourage further investment in domestic alternatives while also accelerating efforts to improve software efficiency and optimise AI systems for available hardware.
At the same time, the restrictions could affect international technology cooperation by increasing uncertainty for companies operating across multiple markets. Countries hosting major data centre facilities may have to balance commercial opportunities with compliance obligations under American export-control regulations.
The broader implications extend beyond the immediate US-China technology rivalry. Artificial intelligence is becoming an important component of economic competitiveness, national security and industrial development. Access to advanced computing infrastructure is therefore increasingly viewed as a strategic resource.
Governments are seeking to secure semiconductor supplies, protect sensitive technologies and maintain reliable access to the hardware required for emerging industries.
The latest American decision signals that export controls are moving beyond conventional restrictions on the physical sale of semiconductor products. Regulators are increasingly examining how advanced technology is accessed, operated and delivered through international digital infrastructure.
As artificial intelligence development continues to accelerate, the ability of governments to monitor computing resources across borders is likely to become an increasingly important part of technology policy.
The effectiveness of the new framework will depend on enforcement, cooperation with international partners and the ability of authorities to identify arrangements designed to bypass licensing requirements.
For the global technology sector, the decision highlights the growing connection between semiconductor manufacturing, cloud computing, artificial intelligence and national security. It also underlines how geopolitical considerations are increasingly shaping the future of the digital economy.