Malaysia's gamble: turning data centres into industrial power

5 January 2026

OpenAI recently wrote that AI has the capacity to “reindustrialise” the United States, and that it aims to invest around USD 1.4 trillion into computing infrastructure over the next decade. Across the world, a parallel version of this story is unfolding — technology firms including Microsoft and Tencent are racing to build data centres abroad, with over two-thirds of data centre capacity under construction in Southeast Asia’s main five economies committed to Malaysia. However, as the country struggles with grid constraints, limited water capacity, and geopolitical pressure from US tariffs, can Malaysia’s industrial strategy convert foreign infrastructure investments into sovereign technological power?

Malaysian AI ambitions, and a national strategy

The Malaysian approach to AI is centred around national economic resilience and digital sovereignty. The government’s 2021 Artificial Intelligence Roadmap emphasises the importance of a national AI ecosystem as “a sine qua non for Malaysia to attain a Developed Nation status by 2030 or even earlier”. Since releasing the Roadmap, Malaysia has risen from 28th to 24th on the Government AI Readiness Index, coming second only to Singapore among ASEAN states.

Malaysia is throwing resources behind its strategy. In a speech on the 2026 state budget, Prime Minister Anwar Ibrahim announced that he would allocate RM 2 billion (around USD 490 million) to building a sovereign AI cloud. Ideally, this would allow AI models to be trained, stored, and deployed on Malaysian land, thus affording Malaysia direct oversight — as opposed to using privately-owned cloud servers managed by foreign corporations. It comes as part of a wider commitment totalling RM 5.9 billion (around USD 1.44 billion) for research, development, commercialisation, and innovation (RDCI) initiatives relating to AI. Combined, the Roadmap and commitment bolster Putrajaya’s vision of AI sovereignty: achieving domestic control over AI development to accelerate economic growth.

Malaysia’s National AI Roadmap reveals tension between its stated ambitions and the industrial reality taking shape. Although Malaysia has courted notable investments into AI datacentres, there is limited focus on cultivating the upstream capabilities that Malaysia identifies as central to its long-term competitiveness.

However, the total budget pales in comparison to international commitments. In January 2025, Trump announced a private sector investment of up to USD 500 billion for AI infrastructure; in November, Meta announced that it would commit over USD 600 billion to the US by 2028. This calls into question the feasibility of plans for sovereign AI, particularly when observed in the context of its limited execution — are aims for sovereign AI too focused on grandiose goals, rather than strategic emphases on specific verticals?

The explicit mapping of AI sovereignty goals has not yet been completed, and the current trajectory of the state’s AI industry seems oriented towards short-term economic gains. In fact, the bulk of Malaysian interactions with AI development comes from foreign investment in data centres funded by multi-national hyperscalers, creating a landscape that risks locking Malaysia into the role of a host economy rather than an active player in the data centre boom.

The data centre boom in Malaysia

Before the COVID-19 pandemic, Asian data centre infrastructure and investments were largely concentrated in established markets, such as Japan, Singapore, and Hong Kong. During the pandemic, accelerated global cloud adoption in Southeast Asia and new restrictions in Singapore — after a three-year moratorium, the state now institutes rigid green data centre specifications — led to significant momentum for data centre development in Malaysia from companies like Google, Amazon, Bytedance, Nvidia, and Alibaba. In addition to lack of land or energy in more developed Asian countries, Malaysia’s proximity to Singapore serves as a major factor for its data centre growth. Malaysia is now gaining data centre capacity faster than any other state in the Asia-Pacific region, with about 850 MW in potential electricity demand announced in the first half of 2024 alone.

However, there is limited research on the types of data centres being developed. Many existing commitments are geared toward inference, or model deployment, and not model training, which would require more computational power and specialised chips. Although these data centres are no doubt economically valuable, they do not meaningfully shift Malaysia into a global group of “compute-north” countries that shape how advanced models are trained and built. Additionally, data centres rarely provide the technological and labour spillover that semiconductor fabrication plants and advanced manufacturing facilities generate.

Workforce adaptation

The Asia-Pacific Data Center Association (APDCA) estimates that the Malaysian AI industry is projected to create 30,900 jobs annually by 2030, a substantial portion of which will be made up of “high-value” roles, which include network engineers, data specialists, cloud infrastructure specialists, and ICT security professionals. APDCA claims that data centre construction alone has contributed USD 24 billion annually in economic output, and that the data centres are expected to deliver USD 10.2 billion in economic output every year. This is from generating investment, supporting highly-skilled jobs, and enabling overall industry growth. However, there is limited transparency related to how data centre operations will translate to the above figure.

This economic output might not translate to tangible, local job creation. Already, onlookers have identified a shortage of skilled labour in Malaysia. Part of this challenge is the timeline of construction. Despite investments into training local Malaysians, the sheer speed at which these projects are moving impedes job training at scale.

To complicate matters, research has shown that data centres create the lowest number of jobs per square foot of facility, employing thousands during construction but barely 200 during operation. Construction could represent high momentum that quickly tapers into minimal operational staffing, therefore offering little in terms of industrial transformation. The jobs created may largely be temporary construction roles and low-level maintenance positions, while research and engineering roles are filled by foreign workers.

Infrastructure strain

Rapid data centre construction has pushed Malaysia’s electricity and water systems to their limits. Water-related shortages in Johor and Selangor have already forced state authorities to slow approvals for data centre construction, as water is widely used for cooling capacity among data centres. This infrastructure strain reveals a troubling asymmetry in dynamics: the Malaysian state is bearing steep environmental costs — a vast majority of the electricity grid in use is powered by coal and gas — while foreign cloud providers access relatively cheap compute infrastructure. Additionally, major technology companies have been acquiring large tracts of land from palm oil plantations to build their facilities, adding a new dimension to the struggles. There is a risk that the country is thus subsidising the computational needs of foreign corporations without ensuring commensurate returns in local innovation capacity or strategic control.

Malaysia is working to address this challenge. First, it plans to add six to eight gigawatts of gas-fired power by 2030 to address growing electricity consumption driven by data centres, but still needs to meet its net zero commitments. Data centres may promise economic growth, but they could also lock the country into carbon-intensive pathways. Second, Malaysia has introduced new power tariffs, which analysts predict could lead to a 10 to 14 per cent increase in power costs for data centre operators. This reflects a tangible response to power capacity strains, but does not resolve the foundational issue of resource-intensive infrastructure expanding faster than the state can adapt.

Geopolitical tensions

Malaysia’s rise as a data centre hub deeply intersects with intensifying US-China competition. As one of China’s largest trading partners in ASEAN, Malaysia faces pressure to maintain economic ties with Beijing, even as US export controls tighten. As AI infrastructure relies on advanced chips, Washington’s restrictions on advanced GPUs — and potential tariffs on data centre equipment — could meaningfully impact Malaysia’s AI build-out, raising costs and limiting access to necessary chips. At the same time, Malaysia’s neutrality has made it an attractive location for companies seeking to diversify supply chains, as well as Chinese businessmen seeking a loophole to circumvent US export controls, in that they were able to access advanced chips through operating in “third-party” nations.

Malaysia has developed a National Semiconductor Strategy to stay in midstream semiconductor production while gradually increasing semiconductor margins and complexity, rather than attempting cutting-edge fabrication. This would position the state as a neutral supplier, taking advantage of current capabilities to incrementally develop products that are higher up the value chain. However, the recent US-Malaysia Reciprocal Trade Deal may complicate this strategy by implicitly aligning Malaysia with US economic priorities.

With an established trade deal with the US, the National Semiconductor Strategy might not achieve the supply chain autonomy that Putrajaya initially envisioned, as compliance with US export restrictions might constrain Malaysia’s ability to serve as a neutral intermediary in the global chip supply chain.

Policy alignment

Malaysia’s National AI Roadmap reveals tension between its stated ambitions and the industrial reality taking shape. Although Malaysia has courted notable investments into AI datacentres, there is limited focus on cultivating the upstream capabilities that Malaysia identifies as central to its long-term competitiveness. Connecting the dots between Malaysia’s attractiveness for foreign tech investment and its national goals should focus on wide-scale adoption, upskilling, and striking careful balances between control and efficiency, foreign tech transfer and indigenous development.

However, these challenges are hardly unique to Malaysia. Many countries are releasing AI policies and strategies based on AI’s potentially transformative effects, while calibrating in real time to the AI ecosystem’s current capabilities and thrust. If Malaysia fails to convert these data centres into engines of local innovation, it risks hosting compute without agency to gain significant value-add. The goal does not necessarily have to be model development or sovereign AI — it is enough if Malaysia can endeavour to capture strategic value at specific points in the AI stack, ensuring that the benefits of the AI revolution are broadly shared.

DISCLAIMER: All views expressed are those of the writer and do not necessarily represent those of the 9DASHLINE.com platform.

Author biography

Faye Simanjuntak is a Schwarzman Fellow at the Asia Society Policy Institute, studying the adoption of AI across Asia. Image credit: Taylor Vick/Unsplash.