Malaysia Faces GDP Hit As Washington’s New AI Chip Licensing Rules Threaten Southeast Asia’s Data Center Pipeline

A potential 0.2 percentage point drag on Malaysia’s GDP growth forecasts through 2028 signals how deeply the country’s economy has become entwined with AI infrastructure investment.

Analysts at Capital Economics warn that a slowdown in data center spending would compound what senior China economist Leah Fahy called “an anticipated cooling in the global AI export cycle,” compressing growth well into the decade.

Malaysia is considered the most exposed country in the region, given the concentration of capacity being built around Johor and the sector’s heavy dependence on Chinese technology clients.

More than half of Malaysia’s data center capacity sits in colocation facilities where Chinese firms account for “a sizeable share of these workloads,” according to Capital Economics, leaving operators vulnerable to the new licensing requirements.

The pressure stems from an interim rule published Monday by the US Commerce Department, which requires data centers in Singapore, Malaysia, Thailand, and Indonesia to obtain government licences before purchasing advanced AI chips, including those made by Nvidia (NVDA).

Facilities that currently serve large Chinese customers have a 90-day window to apply for a licence or risk facing future export restrictions on chip purchases.

The rule formalises Washington’s attempt to sever a supply chain that had quietly taken shape after earlier US restrictions barred Chinese firms from buying Nvidia’s most advanced semiconductors domestically.

Companies including Alibaba Group and ByteDance responded by renting computing capacity from overseas data centres, and October estimates from the Institute for Progress, a Washington-based think tank, projected that roughly 75% of China’s overseas AI compute would be concentrated in Southeast Asia by the close of 2026.

Jessica Brandt, a senior fellow for technology and national security at the Council on Foreign Relations, said the new measures will make building and operating these facilities “slower and costlier,” and warned they could undercut the business model of neoclouds, newer cloud providers built around AI computing, across the region.

Goldman Sachs analysts wrote that Malaysian data centres relying heavily on Chinese customers, particularly those built for AI training, may struggle to lease space and risk being left with idle power capacity, while operators serving US hyperscalers and enterprise clients should remain relatively insulated.

Thailand, which has more than 60 data centres in operation and around 30 under construction, is also bracing for the fallout, with vice minister for commerce Kirida Bhaopichitr warning that new projects failing to secure US chip approval may “not be able to complete construction.”

Governments across the region have largely signalled a willingness to comply, with Malaysia’s investment minister Johari Abdul Ghani stating plainly: “If you want to be in the game, you have to follow U.S. regulations.”

Singapore has taken a different posture, announcing plans to negotiate “trusted access” arrangements for its companies rather than simply accepting the licensing framework as written.

The rule does carve out an exemption for US-headquartered cloud providers, a gap that concerns some national security analysts who fear it creates an alternative route for Chinese firms to access restricted computing power.

Chris McGuire, a senior fellow at the Council on Foreign Relations and former National Security Council official, 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,” and called on Washington to close the loophole before it is exploited.

Reema Bhattacharya, head of Asia research at Verisk Maplecroft, said some high-end capacity could migrate toward Japan, South Korea, and Taiwan, though constraints on land and power supply will limit how much those markets can realistically absorb.

Chinese firms are already searching for workarounds, with developer Moonshot AI reportedly attempting to access Nvidia chips through data centres in Europe, while Oracle scrapped a multibillion-dollar chip-leasing deal with Tencent in early November under pressure from US administration officials, a move that would have provided the Chinese company with roughly 100,000 advanced AI chips.

The rule arrives as Washington continues an investigation into Jensen Huang, who departed from Nvidia, which he co-founded in 1993, following reports of alleged extramarital affairs including one involving a suspected Chinese spy believed to have helped facilitate billions of dollars worth of Nvidia chips reaching China.