New U.S. export rules could slow Southeast Asia's data center boom by tightening access to advanced AI chips and targeting the region's role in providing computing power to Chinese companies.
According to an interim rule released Monday by the Commerce Department, data centers in Singapore, Malaysia, Thailand and Indonesia will need a U.S. government license to purchase advanced AI chips, including those from Nvidia.
Facilities already serving large Chinese customers will have 90 days to apply for a license or face future export restrictions, the rule said.
The region has seen a surge in data center construction to house the semiconductors needed to train and develop artificial intelligence models.
U.S. restrictions on China's access to Nvidia's most advanced chips have pushed Chinese technology firms, including Alibaba Group and ByteDance, to rent computing capacity in overseas data centers.
About 75% of China's overseas AI compute was forecast to be concentrated in Southeast Asia by the end of 2026, according to October estimates by the Institute for Progress, a Washington-based think tank.
The buildout has come under greater scrutiny as Washington presses ahead with the investigation into Nvidia 's former CEO Jensen Huang .
Huang left the company that he co-founded in 1993 after reports of several alleged extramarital affairs, including one with a suspected Chinese spy who is believed to facilitate billions of dollars worth of Nvidia chips into China.
The new restrictions will likely make building and operating these data centers "slower and costlier," said Jessica Brandt, a senior fellow for technology and national security at the Council on Foreign Relations. The rules could undercut the business model of neoclouds — newer cloud providers built around AI computing — in the region, she said.
Operators are likely to face more rigorous regulatory and national security requirements, said Reema Bhattacharya, head of Asia research at risk intelligence company Verisk Maplecroft.
Greater scrutiny over investment into these projects could also weigh on Asian economies that have benefited heavily from the region's data center build-out.
Malaysia stands to lose the most from the new restrictions, analysts say, given the scale of capacity being built in Johor and the sector's reliance on Chinese customers.
More than half of Malaysia's data center capacity is in colocation facilities with Chinese firms accounting for "a sizeable share of these workloads," said Leah Fahy, senior China economist at Capital Economics.
Restricting access to overseas cloud facilities would weaken the commercial case for further expansion and ultimately weigh on investment from both Chinese and U.S. companies, Fahy said.
"Any resulting slowdown in investment would add to the drag from an anticipated cooling in the global AI export cycle," she said, potentially knocking 0.2 percentage points off Malaysia's GDP growth forecasts through 2028.
Thailand, which has more than 60 data centers operating and about 30 under construction, is also bracing for tighter scrutiny.
Thailand's vice minister for commerce, Kirida Bhaopichitr, told CNBC's JP Ong in Manila on Sunday that the country would scrutinize data centers importing high-end chips and "crack down" on those that re-export them to third countries.
New projects that fail to secure U.S. approval for advanced chips may "not be able to complete construction," the vice minister said, though existing facilities providing AI services could be less affected.
Governments in the region have signaled compliance with the U.S. requirements. "We will have to discuss with the U.S. to make sure that the chips are coming," and are allowed for use in data centers operating in Thailand, Bhaopichitr said.
Malaysia's investment, trade and industry minister, Johari Abdul Ghani, said Tuesday, "If you want to be in the game, you have to follow U.S. regulations."
Singapore, meanwhile, has said it plans to negotiate "trusted access" to advanced AI chips for its companies.
The rule, however, exempts U.S.-linked firms, leaving a narrow route for Chinese companies to continue renting advanced computing power through American cloud giants, analysts said, though they might exercise caution.
"Potentially, U.S. firms may receive an advantage," Chong said.
"There is a significant risk that Chinese companies will pivot from renting compute from data centers in Southeast Asia, to renting compute from data centers in Virginia," said Chris McGuire, a senior fellow at the Council on Foreign Relations and former National Security Council official. He urged Washington to close that loophole before it is exploited.
Bhattacharya said some high-end capacity may also shift to Japan, South Korea and Taiwan, though constraints on power and land may restrain how much those markets can absorb.
Chinese technology companies have already begun looking for alternative sources of computing power outside of Southeast Asia. CNBC reported earlier this month that Chinese developer Moonshot AI tried to access Nvidia chips through data centers in Europe.
Oracle canceled a multibillion-dollar chip-leasing agreement with Tencent in early November that would have given the Chinese company access to about 100,000 advanced AI chips, according to the Financial Times , after concerns were raised by U.S. administration officials.
— CNBC's Evelyn Cheng contributed to this report.