By Sara Loo, Associate Research Officer at ISEAS-Yusof Ishak Institute

In May 2025, Malaysia’s Deputy Minister of Communications, Teo Nie Ching, announced the launch of the country’s sovereign full-stack Artificial Intelligence (AI) infrastructure—the first in Southeast Asia. The initial announcement, which mentioned the deployment of Huawei’s advanced Ascend chips as the backbone of a national AI initiative by 2026, sparked controversy amid ongoing US-China tensions. The government later backtracked on the project, asserting that it remains “fiercely independent” in its choice of technologies. As a result, some observers have concluded that sovereign AI is a “fraught pursuit” for countries like Malaysia, which have to lock in external reliance on the US or China for advanced chips.

Beyond advanced chips—which, in today’s geopolitical context, require making choices that imply allegiance to a particular brand and, by extension, a country—there is much more to sovereign AI that needs to be unpacked. What is sovereign AI? Why do countries like Malaysia claim it to be important? Are there opportunities for these countries to lead on AI sovereignty given external reliance in its AI supply chain? External reliance for advanced chips may be a given. But Malaysia still has the potential to take the lead in standard setting for governance surrounding data, infrastructure and talent to emerge as a leading digital powerhouse.

What is Sovereign AI and Malaysia’s Focus

Sovereign AI has been a core tenet of Malaysia’s AI policy and is one of the seven working groups under the National AI Office (NAIO), an office launched in December 2024 to spearhead AI development and integration across various sectors. Broadly speaking, AI sovereignty refers to a nation’s ability to use its own resources to shape the development of its AI, maintaining control over AI capabilities, data, and infrastructure. Conversations with policymakers and AI startups in Malaysia point to three key pillars of sovereign AI: local data, local infrastructure, and local talent. In the case of a large language model (LLM), a sovereign AI system would involve collecting data in Malaysia, training it using domestic infrastructure such as in data centres in the country, and employing local talent, such as generative AI developers and engineers. In a press release in April this year, the Ministry of Digital also affirmed that “talent development” and “data centre growth” are means to build a “robust AI ecosystem”. The three pillars are unpacked below.

Pillar One: Local Data

Local data can be understood as the “fuel” for sovereign AI. A robust data governance framework tailored to specific AI use cases would be critical for any sovereign AI model.

Malaysia has made progress on the data governance front over the past few years. For example, it made amendments to the Personal Data Protection Act (PDPA) 2010 in July 2024. The amendment better aligns PDPA with international frameworks like EU’s GDPR and introduced a direct obligation for data processors including requiring them to take practical steps to safeguard personal data. Additionally, the amendment creates a more flexible cross-border data transfer regime, allowing cross-border data transfers outside Malaysia if that place has laws substantially similar to the PDPA or provides an adequate level of protection in relation to the processing of personal data at least equivalent to the level of protection afforded by the PDPA.

In December 2024, Malaysia also passed a law that allows data sharing between federal government agencies, decided by the national data sharing committee. The Ministry of Digital emphasised that the focus is on ensuring data is secure in the sharing process.

While these efforts are not targeted at AI specifically, they lay the groundwork for potential AI-focused data governance laws.

Pillar Two: Local Infrastructure

While advanced chips are a critical part of infrastructure—without which AI systems would not be able to operate—it is far from the only hardware required. Malaysia has in fact been building up critical infrastructure that serve as the backbone of AI in the country, effectively expanding local AI computing power. This has come in the form of building up its data centre capacity, with a focus on “AI data centres” since at least September 2024. Recent estimates suggest that Malaysia’s data centre market was valued at USD 4.04 billion in 2024 and is projected to reach USD 13.57 billion by 2030. The country, specifically the state of Johor, is widely described as the fastest growing data centre market in Asia Pacific.

Such rapid expansion in infrastructure comes with risks, however. The country has come under increasing scrutiny surrounding its lack of enforceable regulations for the rising demand for energy and water. In response, national utilities company Tenaga Nasional Berhad (TNB) introduced a band pricing under which hyperscale data centres would be expected to fall in the ultra-high voltage category with the highest tariffs. This pricing policy has, however, been criticised as “unclear” and inducing “uncertainty”. There have also been similar discussions surrounding hikes in water tariffs and calls for more stringent water use regulations. These discussions have yet to produce enforceable guidelines, but it remains another area where Malaysia, given its current status as a leading data centre market, has the potential to take the lead in standard setting to establish itself as a leading tech powerhouse that prioritises sustainable growth.

Pillar Three: Local Talent

Malaysia has long been plagued by brain drain. In the face of AI, new skills are required to equip its population to take on new jobs. A study by TalentCorp found that approximately 620,000 jobs are at high risk of being replaced or becoming obsolete due to automation, and there would be hundreds of thousands of 60 emerging roles, with 70% of these roles being in AI and digital technologies. This transformation in the demands of the workforce requires cross-skilling, upskilling and reskilling.

To this end, Malaysia has launched several initiatives to strengthen its AI workforce. In May, NAIO and TalentCorp announced the formation of the MyMahir National AI Council to accelerate Malaysia’s AI workforce readiness.

Industry players also play an important role in building up local talent readiness. In this author’s interview with Khalil Nooh, co-founder of Mesolitica, a Malaysia-focused large language model startup, he emphasised the importance of “making genAI courses local”. Noting that there is no shortage of education or content from big tech companies, Khalil said that his team has been coming up with their own syllabus and training in response to “a lot of requests for training” since the end of 2024. While such up- and cross-skilling will take years and Malaysia appears to be at an early stage still, Khalil also highlighted that the few years taken for the first batch of graduates from new AI syllabus would correspond with the new data centres coming online in about three to four years.

Purported Benefits of Sovereign AI

Malaysia is not the only country to emphasise sovereign AI. Around the world, countries are prioritising it for both economic and national security reasons. Economically, developing a local talent pool with the knowledge and skills to harness AI can stimulate growth and position countries as leaders in the global digital economy. Doing so allows countries to avoid building a digital economy that “exports little, owns less and pays more”, driven by rising licensing fees, AI subscriptions and reliance on foreign cloud infrastructure. From a national security perspective, sovereign AI helps protect against potential supply chain disruptions, thereby reinforcing national sovereignty. In terms of privacy, using domestic servers and localising LLMs like DeepSeek means data will be processed locally, which government officials claim will safeguard user privacy and strengthen data security.

Looking Forward: Policy Priorities

A robust sovereign AI framework requires not only domestic capability but also readiness for cross-border collaboration and investment in talent development. Sound data-sharing practices and forward-looking talent policies are critical to advancing AI sovereignty.

As CEO of NAIO, Sam Majid recognised in an interview with this author, Malaysia “can share data and lessons learned with another vertical in another country…among ASEAN countries”, recognising that one of NAIO’s KPIs is to provide clarity on how data is shared when it comes to AI use cases.

With less than four months remaining in NAIO’s one-year incubation period, Malaysia should seize the opportunity to lead in shaping AI governance guidelines that are applicable across ASEAN. Taking the initiative to develop updated data-sharing frameworks specifically tailored to AI use cases could serve as a model for AI sovereignty.

The views and recommendations expressed in this article are solely of the author and do not necessarily reflect the views and position of the Tech for Good Institute.