Skip to main content
  • Home
  • Global Tech
  • “Is the Investment Boom Cooling?” Southeast Asian Data Center Stocks Plunge as U.S. Export Controls and Local Environmental Rules Fuel AI Infrastructure Bubble Fears

“Is the Investment Boom Cooling?” Southeast Asian Data Center Stocks Plunge as U.S. Export Controls and Local Environmental Rules Fuel AI Infrastructure Bubble Fears

Picture

Member for

2 years
Real name
Tyler Hansbrough
Bio
[email protected]

As one of the youngest members of the team, Tyler Hansbrough is a rising star in financial journalism. His fresh perspective and analytical approach bring a modern edge to business reporting. Whether he’s covering stock market trends or dissecting corporate earnings, his sharp insights resonate with the new generation of investors.

Modified

Southeast Asian data center and semiconductor stocks plunge, exposing the downside of massive investment inflows
U.S. tightens advanced chip export controls, cutting off Chinese companies’ access to computing through third countries
Concerns mount over overinvestment and profitability in the global AI infrastructure market

Shares in major Southeast Asian data center and semiconductor companies fell across the board. Investor sentiment weakened sharply as growing local opposition to data centers’ enormous electricity and water consumption coincided with expanded U.S. export controls on advanced artificial intelligence (AI) chips and data center computing resources. Market observers warn that this trend could spread beyond Southeast Asia to the global AI infrastructure market, deepening doubts about the profitability of data centers themselves.

Cracks Emerge in Southeast Asia’s Data Center Investment Boom

According to Reuters on the 17th (local time), shares in major Southeast Asian data center and semiconductor supply chain companies broadly declined that day. In Malaysia, shares in data center infrastructure companies MN Holdings and Critical Holdings plunged nearly 20%, while YTL Power, which collaborates with Nvidia on AI infrastructure, fell about 10%. Semiconductor-related stocks also showed pronounced weakness in Singapore. Frencken, which supplies components to Dutch semiconductor equipment maker ASML, fell 3.91%, while AEM, a semiconductor testing equipment supplier that does business with Intel, declined 2.9%.

Several factors lie behind this sharp deterioration in investor sentiment. First, local market perceptions of the AI infrastructure industry are changing rapidly. Southeast Asian countries have attracted massive investments from global technology giants as alternative destinations to established data center hubs such as the United States and Singapore. Malaysia’s Johor state secured USD 35 billion in data center investments from companies including Amazon, Microsoft (MS), Tencent and Alibaba, while Thailand also saw a surge in investment applications centered on AI infrastructure. According to Thailand’s Board of Investment (BOI), total investment applications from January through September this year rose 28% from a year earlier to USD 50.3 billion. AI supply chain investments, including data centers, cloud services, and semiconductor and server manufacturing, accounted for about 70% of that total.

Environmental Rules Tighten as Local Opposition Spreads

The problem is that local opposition has recently been spreading over data centers’ enormous electricity and water consumption and environmental damage. In February, for example, residents of Iskandar Puteri in Johor protested a 300-megawatt (MW) data center project proposed by Chinese-owned company ZDATA. More recently, dozens of residents also opposed data center construction at Nongsa Digital Park on Indonesia’s Batam Island. As a prolonged drought and interruptions to tap water supplies left residents struggling to cook and bathe, complaints emerged that data centers’ use of cooling water was worsening the shortage.

Southeast Asian governments have also begun moving beyond their single-minded pursuit of investment to tighten environmental and resource regulations. On the 10th, the Thai government approved new standards limiting noise at data center site boundaries to 70 decibels (dBA) and requiring increases in surrounding temperatures to be kept within 1–2 degrees. It also mandated cooling-water recirculation and testing before discharge, and introduced a requirement that data centers consuming 20 MW or more of electricity be built within industrial estates. Malaysia’s government is likewise raising the approval threshold for data center projects. Last month, Malaysian Minister of Investment, Trade and Industry Johari Abdul Ghani said new data center construction would not be approved in areas without sufficient electricity and water supplies.

Table 1. Factors Weakening Investor Sentiment Toward Southeast Asian AI Infrastructure

CategorySpecific FactorsKey Details
Industry ExpansionConcentration of Large-Scale InvestmentSurging investment in data centers, cloud services and semiconductors raises concerns about infrastructure capacity and business sustainability
Environment and ResourcesLocal OppositionGrowing community opposition over data centers’ electricity and water consumption and environmental damage
Government RegulationNew environmental standards and stricter reviews of new projects increasingly constrain related business development
U.S. Export ControlsChip and Computing RestrictionsPrior authorization required for AI chip exports to key Southeast Asian countries; controls on remote provision of computing resources to Chinese customers
Sources: Reuters, Thailand’s Board of Investment (BOI), Malaysia’s Ministry of Investment, Trade and Industry, U.S. Department of Commerce’s Bureau of Industry and Security (BIS)

U.S. Chip Export Controls Take Direct Aim at Southeast Asia

Recently announced U.S. semiconductor export controls have intensified this market downturn. On the 15th, the U.S. Department of Commerce’s Bureau of Industry and Security (BIS) designated five countries—Singapore, Thailand, Malaysia, Indonesia and Brazil—as presenting a high risk of illegal diversion of advanced chips. It issued an interim final rule (IFR) requiring prior U.S. government authorization for exports, reexports and in-country transfers of advanced AI chips to or within these countries. Separate export licensing requirements also apply to overseas data centers that remotely provide large-scale computing capacity to companies headquartered in countries subject to U.S. export controls, such as China. Data center operators and chip suppliers must verify customers’ locations and their ultimate ownership and control.

These measures come as Chinese companies use Southeast Asia as a base for circumventing U.S. trade restrictions. Unable to directly import Nvidia’s advanced AI chips, Chinese companies have been developing AI models by contracting with cloud providers in third countries such as Malaysia and Singapore and renting the computing capacity of chips installed there. In one example, The Wall Street Journal (WSJ) reported in June last year that four Chinese engineers brought about 80 terabytes (TB) of training data into Malaysia on 60 hard drives and trained an AI model using about 300 locally rented servers equipped with Nvidia chips.

China’s Strategy for Securing Computing Resources Through Third Countries

Chinese companies have also poured substantial sums into the region. In 2024, ByteDance announced plans to establish a USD 2.13 billion AI hub in Malaysia and invest an additional USD 320 million to expand its data centers in Johor. In February last year, it also announced plans to invest USD 8.8 billion in Thai data centers over the following five years. Since then, ByteDance has reportedly been working this year with Southeast Asian cloud provider Aolani Cloud on a plan to install about 500 Nvidia Blackwell computing systems in Malaysia. The facilities would contain approximately 36,000 B200 chips, with construction costs estimated to exceed USD 2.5 billion.

China’s indirect access to computing resources has come under scrutiny following recent controversy surrounding former Nvidia Chief Executive Officer (CEO) Jensen Huang. At the center of the controversy is Singaporean cloud company Megaspeed, where Alice Huang, a person with a private relationship with the former CEO who is suspected of ties to Chinese intelligence agencies, previously served as an executive. The U.S. Department of Commerce began investigating Megaspeed in 2024 after detecting signs of illegal diversion of advanced chips, and Nvidia halted chip sales to the company in June last year. It subsequently emerged that Megaspeed’s Malaysian subsidiary, Speedmatrix, had pledged servers and networking equipment as collateral to Alibaba subsidiary Apex Enterprise Solutions, raising suspicions that Chinese companies were providing financing for purchases through intermediaries.

Returns on AI Investment Remain Uncertain

Some observers predict that this weakening of sentiment toward data center investment could spread beyond Southeast Asia to the broader global AI market. They argue that the current situation will deepen investors’ already growing doubts about data centers’ ability to generate revenue. In September, global consultancy Bain & Company estimated that worldwide AI infrastructure investment would reach USD 1.5 trillion annually by 2031, and that the AI industry would need to generate USD 6 trillion in annual revenue to justify it. Expected revenue from existing AI services amounts to no more than USD 1.8 trillion, leaving the remaining USD 4.2 trillion to come from new businesses such as robotics, automation and drug discovery.

Overestimates of data center development plans and electricity demand are also weighing on investor sentiment. In September, Reuters reported that requests for large-scale power supplies, primarily for data centers and similar facilities, had exceeded 700 gigawatts (GW) across major U.S. regions, but that many were duplicate applications or came from projects with uncertain financing. U.S. utility Exelon cut its estimate of likely data center electricity demand by about 40% after tightening developers’ collateral requirements. “If data center development plans unsupported by genuine demand begin to be weeded out in earnest, enthusiasm for AI infrastructure investment could cool rapidly,” one market expert said. “As the problem of so-called ‘ghost data centers’ with little prospect of materializing becomes apparent, the profitability of similar projects and the likelihood of recouping investments will be reassessed.”

Picture

Member for

2 years
Real name
Tyler Hansbrough
Bio
[email protected]

As one of the youngest members of the team, Tyler Hansbrough is a rising star in financial journalism. His fresh perspective and analytical approach bring a modern edge to business reporting. Whether he’s covering stock market trends or dissecting corporate earnings, his sharp insights resonate with the new generation of investors.

Similar Post