The Trump administration is reportedly drafting a new rule that would restrict shipments of US-made AI chipsets to Malaysia and Thailand in an effort to prevent the chips from finding their way into China.
According to a report from Bloomberg on Friday, citing “people familiar with the matter”, the draft rule from the Commerce Department would update the US government’s current “AI diffusion rule that was initiated under President Joe Biden at the end of 2024 to restrict AI chip exports to countries suspected by the US government of serving as intermediaries to smuggle the chips to China.
The Commerce Department announced in May it would rescind and replace the AI diffusion rule, which it said was too sweeping in scope. The reported draft rule isn’t a replacement so much as a first step towards a new policy that would leave heavy export controls restrictions in place for countries including China, Cuba, Iran, North Korea and Russia, while adding Malaysia and Thailand to the list, the report said.
Sources also told Bloomberg the draft rule isn’t finalised and could still change.
Nvidia CEO Jensen Huang – who opposed the Biden administration’s AI diffusion rule – has been previously quoted as saying there’s “no evidence” of AI chips being diverted to countries on the Commerce Department’s heavy exports control list.
The impact of the rule as reportedly written for Malaysia and Thailand is unclear. The rule includes exemptions that would, for example, enable businesses headquartered in the US and some allied nations to keep shipping AI chips to both countries for a few months after the rule is published, the report said.
According to The Edge Malaysia, Malaysian stocks for tech and construction companies involved in data centres fell on Monday morning as a result of the Bloomberg report, along with investor uncertainty as the country faces a July 9 deadline for a US tariff deal.
Both Malaysia and Thailand are building themselves up as major AI data centre hubs in Southeast Asia.
Nvidia already has several chipset customers in Malaysia, including Telekom Malaysia, whose recently announced GPU-as-a-Service uses Nvidia GPUs, and YTL Power International, who is collaborating with Nvidia to build an AI data centre in Johor.
In Thailand, cloud-computing company Siam.AI Cloud is developing a Thai large language model (LLM) using H100 Tensor Core GPUs in partnership with South Korean company Naver.
