How new US chip restrictions on South East Asia will hit China
The new interim US rules increase scrutiny of data centre operators in Singapore, Malaysia, Thailand, Indonesia and Brazil. They are intended to stop advanced US chips, and the computing power those chips provide, from reaching Chinese firms through overseas facilities. This matters because Chinese companies rent computing services abroad to train AI models. Operators and chip buyers face closer checks on who will use the chips and computing services. Firms will need a US government licence to buy advanced chips from Nvidia or other American chipmakers. The US says this will help it monitor whether chips serve Chinese companies or are being smuggled to China. The Bureau of Industry and Security says the five countries present a higher-than-normal risk of chip diversion. It says licences will be issued case by case, with the goal of preventing circumvention through remote computing arrangements. Singapore says it is engaging US authorities and exploring trusted access for local firms. The rules could make it harder for some regional data centres to obtain advanced chips.
What do the new US rules require of data centre operators in Singapore, Malaysia, Thailand, Indonesia and Brazil?
The new interim US rules increase scrutiny of data centre operators in Singapore, Malaysia, Thailand, Indonesia and Brazil. They are intended to stop advanced US chips, and the computing power those chips provide, from reaching Chinese firms through overseas facilities. This matters because Chinese companies rent computing services abroad to train AI models.
Operators and chip buyers face closer checks on who will use the chips and computing services. Firms will need a US government licence to buy advanced chips from Nvidia or other American chipmakers. The US says this will help it monitor whether chips serve Chinese companies or are being smuggled to China.
The Bureau of Industry and Security says the five countries present a higher-than-normal risk of chip diversion. It says licences will be issued case by case, with the goal of preventing circumvention through remote computing arrangements. Singapore says it is engaging US authorities and exploring trusted access for local firms. The rules could make it harder for some regional data centres to obtain advanced chips.
How do the rules restrict Chinese firms from using advanced chips through computing services rented overseas?
The new rules address a route Chinese firms have used to access advanced computing: renting capacity from data centres abroad. This matters because US export restrictions have left China more dependent on overseas compute, which companies use to train AI models. Washington says sensitive US technology should not be reached through remote computing arrangements in third countries.
The mechanism combines licensing and scrutiny. Companies need a US government licence to buy advanced chips from Nvidia or other American chipmakers. US authorities can then monitor whether those chips are used to serve Chinese firms or are diverted to China. The rules increase scrutiny of data centre operators in Singapore, Malaysia, Thailand, Indonesia and Brazil, and licences will be reviewed case by case.
The rules could make overseas computing harder for Chinese firms to obtain, but the article does not say all access will end. The Institute for Progress projects that at least 30% of China’s advanced AI computing power could come from overseas rentals or smuggling in 2027. Beijing called the measures economic coercion and said it would take necessary measures.
What share of China’s advanced AI computing power could come from overseas rentals or smuggling in 2027?
The Institute for Progress projects that at least 30% of China’s advanced AI computing power could come from overseas rentals or smuggling in 2027. The estimate signals how important access beyond China’s borders may remain for advanced AI work. The article links this reliance to US export restrictions, which have made China more dependent on computing capacity abroad.
Renting overseas compute lets Chinese firms use the processing capacity of data centres without necessarily bringing the chips into China. The article says Chinese companies currently rent computing power abroad to train AI models. The new US rules increase scrutiny of facilities in Singapore, Malaysia, Thailand, Indonesia and Brazil, where authorities say there is a higher-than-normal risk of chip diversion.
The 30% figure is a projection, not a report of the share already being used. The restrictions could affect this source of capacity by requiring US licences for advanced American chips and checking whether they serve Chinese firms. The article does not quantify how much the rules will change the 2027 estimate. China has called the measures economic coercion.
Why have Southeast Asian countries become important locations for data centres serving Chinese technology firms?
Southeast Asian countries have invested heavily in data centres to support their AI ambitions. These facilities can provide computing power to companies that need to train AI models. Chinese firms seeking that capacity overseas have been an important market, especially as US export restrictions have made China more dependent on compute from abroad.
The region’s growth also reflects a shift in where projects are built. Singapore was an early adopter, but when it began limiting new construction, investment flowed to neighbours Malaysia and Thailand. Johor in Malaysia and Batam in Indonesia approved billions of dollars worth of data centres to capture spillover demand. Countries have offered incentives such as cheaper land, including to Chinese technology giants Alibaba and Tencent.
This combination of infrastructure, demand and investment has made the region important to firms seeking computing services. But the same connections have drawn US scrutiny: authorities say the five countries present a higher-than-normal risk of chip diversion. New licensing requirements could make advanced chips harder for regional data centres to obtain, particularly where Chinese buyers have driven rollouts.
What effects could the restrictions have on Chinese AI development and data centre investment in Southeast Asia?
For Chinese firms, the rules could restrict access to advanced chips and to computing power rented from overseas data centres. That matters because China has become deeply dependent on compute from abroad following US export restrictions, and companies use it to train AI models. The Institute for Progress projects at least 30% of China’s advanced AI computing power could come from overseas rentals or smuggling in 2027.
The regional impact could fall unevenly. International trade expert Naoise McDonagh says Malaysia and Thailand may be hit hardest because Chinese buyers have heavily driven their data centre rollouts. The new rules require US licences for advanced chips and add scrutiny to remote computing, aiming to stop technology from reaching China through third-country facilities.
The article does not say the restrictions will halt investment or Chinese AI development. Reema Bhattacharya of Verisk Maplecroft says the data centre boom as a whole is unlikely to be derailed. Investment backed by major technology firms and customers that can be clearly identified should remain relatively resilient. Singapore is engaging US authorities, while Malaysia says it will not allow export controls to be circumvented.
Why does China still need access to overseas computing power despite making technology self-sufficiency a national priority?
Beijing has made technology self-sufficiency a national priority, but the article says China has become deeply dependent on compute from abroad after US export restrictions. The gap between a policy goal and current access matters because Chinese firms need powerful computing capacity to train AI models and develop advanced technology.
The article says Chinese companies rent computing power overseas, and estimates that at least 30% of China’s advanced AI computing power could come from overseas rentals or smuggling in 2027. It also says Beijing needs the best chips to build powerful AI models. The US rules are designed to limit access to advanced chips and prevent their computing power reaching Chinese firms through third-country data centres.
Chinese firms Pramana and MoonShot have come close to US rivals Paradox and Axiom, according to the article. The Institute for Progress estimates that top US AI models lead the best Chinese platforms by about six and a half months. That reported progress does not remove the need for computing resources; the new licensing rules create another challenge for access from abroad.
What is a semiconductor, and how do advanced chips help power artificial intelligence?
A semiconductor is a material that conducts electricity less freely than a conductor such as copper, but more freely than an insulator. Its electrical behaviour can be controlled, which makes it useful for building electronic components. This is general background knowledge; the article does not define semiconductors.
A chip is a small electronic device containing circuits made from semiconductor materials. Advanced chips can perform large numbers of calculations quickly. In data centres, groups of chips provide computing capacity that companies can rent. The article’s example is Chinese firms using rented overseas computing power to train artificial intelligence models.
The new US rules matter because they restrict access to advanced chips from Nvidia and other American chipmakers, as well as scrutiny of the computing power those chips provide. Companies need a US government licence to buy them. The article says the rules aim to prevent sensitive US technology from being accessed through remote computing arrangements involving data centres in third countries. The precise effect will depend on case-by-case licence decisions.
What does “compute” mean when Chinese firms rent computing services from overseas data centres?
Here, “compute” means computing capacity: the ability of computers and chips to process information and run demanding tasks. The article uses the term for computing power that Chinese firms rent from data centres abroad, especially to train AI models. It is not a separate product from the facilities; it is the service those facilities provide using their computing equipment.
For example, a Chinese company can rent computing services from a data centre in another country instead of relying only on chips located in China. Advanced chips, including those made by Nvidia and other American chipmakers, supply much of the capacity at issue. The new US rules increase scrutiny of whether such chips are used to serve Chinese firms through remote computing arrangements.
This access matters because US export restrictions have made China deeply dependent on compute from abroad. The Institute for Progress projects that at least 30% of China’s advanced AI computing power could come from overseas rentals or smuggling in 2027. US licences will be decided case by case, so the rules may limit some access without the article claiming that all overseas compute will stop.
This brief was written by AI from the original reporting and checked by other models. Names, figures and quotes come from the source; read it for full context.
Read more in the JupiteX app
Pulse is free. New stories every 4 hours, each one broken into the questions that explain it.
Or read more news on the web