US imposes new export curbs on AI chips for five nations
The US is tightening controls on high-end AI hardware to prevent smuggling and remote computing access by Chinese firms.

The short version
- New rules target five nations deemed at "higher risk" for diverting advanced AI chips to China.
- Shipments of covered chips to these countries now require a specific US export license.
- Data centers providing over 12 million aggregate total processing performance (TPP) face new requirements.
- The policy aims to close loopholes where firms rent remote computing power instead of buying chips directly.
- US-owned data centers and social media platforms are largely exempt from these new restrictions.
Why it matters
This move signals a shift from just blocking hardware sales to blocking the 'renting' of computing power. It makes it harder for international firms to bypass trade barriers by using third-party data centers.
New restrictions on five nations
The United States government has implemented stricter export controls on advanced artificial intelligence (AI) chips. These measures specifically target five nations: Malaysia, Singapore, Thailand, Indonesia, and Brazil. The US Commerce Department identified these countries as locations where high-end chips are at a "higher risk" of being diverted to China.
The primary goal of this policy is to crack down on the smuggling of high-end hardware. By tightening these rules, the US aims to limit the ability of Chinese entities to acquire the advanced technology needed for sophisticated AI development. The move follows reports of illicit procurement and potential re-export of Nvidia chips from certain regional partners.
Under the new guidelines, shipments of covered chips to any of these five countries now require a formal US export license. This adds a layer of bureaucratic oversight for companies moving hardware across borders. The government is specifically looking to prevent these chips from ending up in the hands of restricted foreign actors.
Targeting remote computing power
A major component of the new rules involves overseas data centers. The US is looking to stop firms headquartered in China from renting significant volumes of computing power from facilities located in these five countries. This is intended to close a loophole where companies cannot buy chips directly but can still access the power via remote servers.
The government has set a specific threshold for these restrictions. Only data centers offering a combined processing capacity exceeding 12 million aggregate total processing performance (TPP) will be subject to these new rules. TPP is a metric used to measure the total amount of computing work a system can perform at once.
By focusing on high-capacity facilities, the US hopes to limit the scale of AI development available to Chinese firms. This strategy addresses the reality that many companies prefer to rent cloud capacity rather than manage their own physical hardware. The new rules ensure that large-scale remote access is subject to the same scrutiny as physical exports.
Manufacturing and packaging limits
The US Commerce Department also tightened restrictions on overseas manufacturing and packaging arrangements. This means that chip designers who are affected by the new rules face limitations on using foreign foundries without explicit US approval. The move seeks to control the entire lifecycle of advanced chip production, not just the final sale.
These rules are designed to prevent the assembly of advanced components in regions that might facilitate unauthorized access. By requiring approval for manufacturing steps, the US can better track where the technology is being processed. This creates a more rigorous environment for companies involved in the global semiconductor supply chain.
The policy reflects a broader strategy to secure the supply of advanced technology. It ensures that the infrastructure used to build and package these chips remains under the oversight of the US government. This is particularly important for high-end chips produced by major companies like Nvidia.
Who is exempt from the rules
Not all entities operating in the five target countries will be affected by these new restrictions. The US government has clarified that these rules do not apply to US companies that operate their own data centers in Malaysia, Singapore, Thailand, Indonesia, or Brazil. These companies can continue their operations provided they meet specific reporting requirements.
Additionally, systems used by online retailers and social media companies are currently exempt from these measures. The government determined that these specific types of services do not pose the same level of national security threat as high-end AI computing. This distinction allows common internet services to function without additional hurdles.
The impact on specific countries like Malaysia will depend heavily on the ownership of the companies involved. As reported by Free Malaysia Today on MSN, US cloud providers bringing chips in for their own use may qualify for exceptions. However, local operators receiving chips directly may face more significant licensing hurdles.
The path forward for data centers
Malaysian authorities have already begun investigating specific companies over allegations of illicit procurement. One such firm, Speedmatrix Sdn Bhd, is being looked into due to its links to a Singapore-based cloud company. The investigation centers on the suspicion that high-end Nvidia AI hardware might be illegally moved or sent to China.
Data center operators in the region will need to be more diligent about their customer base. Facilities that provide massive computing resources to clients in prohibited regions, such as China, Hong Kong, and Macau, will face increased scrutiny. They will likely need to provide more transparency regarding who is actually using the computing power they provide.
Moving forward, the complexity of the global chip trade will continue to increase for regional players. Companies must navigate a landscape where the ownership of the end-user is just as important as the physical location of the hardware. This shift marks a new era of enforcement where the flow of data and computing power is monitored as closely as the hardware itself.
What to watch
- How many data centers will actually meet the 12M TPP threshold?
- Whether more countries will be added to the 'higher risk' list in the coming months.
- The results of the investigation into Speedmatrix Sdn Bhd and its partners.
Try it
Check if your company uses cloud computing services from providers based in Malaysia or Singapore to see if they fall under the new reporting requirements.
Sources and links
First reported by Free Malaysia Today on MSN. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.