To this end, governments worldwide have begun articulating national AI strategies that increasingly place AI sovereignty at their core. By 2028, Gartner predicts that 65% of governments worldwide will have adopted some form of AI sovereignty to limit dependency on external parties. This marks a notable departure from an earlier mindset in AI adoption, which largely viewed it through a Software as a Service (SaaS) lens, emphasising convenience and cost-effectiveness. Today, the focus instead lies on safeguarding national interests and ensuring that critical AI capabilities can be developed and governed domestically — on concerns of AI sovereignty.
Sovereignty defined
AI sovereignty refers to a nation’s ability to independently develop, deploy and govern AI with minimal reliance on foreign actors. It encompasses control over systems, data and decision-making processes, and touches on every layer of the ecosystem — from the location of data centres and data flows to the ownership, governance, operation and provision of the hardware, software and services that power them.
See also: The differences between crypto, stablecoin and CBDC
In practice, AI sovereignty aligns with the broader “AI stack”, the ecosystem of interdependent technologies needed to bring AI systems to life. Typically, it is structured into three to six layers. For the purpose of national strategy, the stack can be distilled into three critical layers: compute, data and models. Additionally, achieving AI sovereignty also requires two supporting pillars — talent and regulatory oversight — that act as key enablers for the effective and sustainable deployment of sovereign AI systems.
Sovereignty as it pertains to each dimension is defined as follows:
Compute sovereignty: Control over the computing infrastructure that powers AI, spanning energy supply, semiconductor manufacturing, graphics processing units (GPUs) and data centre assets.
See also: Crypto investing as an exit asset: Save the evangelism
Model sovereignty: Control over the development, training and/or meaningful adaptation of AI models for the purpose of retaining strategic autonomy over their performance, security and alignment over their life cycle.
Data sovereignty: Control and governance of data generated, stored and processed within national borders.
Talent: Access to a robust domestic workforce capable of building, managing and maintaining AI systems (for example, researchers, engineers, data scientists and policy experts), supported by institutions that cultivate such talent.
Regulatory oversight: Comprehensive security, safety and ethical standards to address the rapidly evolving risks and technological developments in AI.
Selective competence
To be clear, AI sovereignty is not about isolation but strategic independence. It differs from the foolhardy pursuit of technological autarky, which is neither feasible nor desirable in a world of globally integrated supply chains and digital interconnectedness. Even the most advanced powers face inherent limitations in achieving full-stack self-sufficiency: the US navigates energy bottlenecks and fragmented regulation, while China grapples with constraints in high-end manufacturing and talent.
These challenges are even more pronounced for the rest of the world, especially smaller economies with limited resources and capabilities. This, in turn, raises a pressing question: What should smaller nations do?
For more stories about where money flows, click here for Capital Section
As wisdom distilled from centuries of military and strategic thinking suggests, pick your battles.
Sovereignty, as defined earlier, is not a single capability but a layered condition. In practice, countries rarely command every aspect of the AI stack, operating instead within a patchwork of strengths and dependencies across the given layers outlined above. The goal is not to secure a foothold in every layer of the stack. Rather, the policy challenge lies in identifying strategic priorities and securing control over them.
The real question surrounding sovereignty, then, is one of strategic intent: How far, and along which dimensions, should a country pursue sovereignty?
Practical implementation
Once the illusion of competing across the entire stack is abandoned, three broad policy options emerge. For any given layer, a country may choose to build capabilities domestically, buy them from abroad or partner with foreign providers to meet its AI needs. The choice depends on how each AI layer relates to national priorities and where the country has a comparative advantage, allowing it to focus on areas of greatest strategic impact.
On a practical level, multiple pathways exist for asserting sovereignty. Countries worldwide have adopted vastly different approaches. At one end of the spectrum, China and the UAE pursue strongly state-led strategies, whereas the US leans more heavily on private-sector dynamism, complemented by targeted state support through hybrid initiatives (for instance, the Stargate Project).
In all cases, sovereign frameworks remain fundamentally shaped by negotiations between governments, corporations and other actors. But regardless of the approach chosen, one thing remains clear:Effective sovereign AI strategies require active state involvement to ensure that AI development aligns with national interests.
What might sovereignty look like?
Malaysia stands at a critical juncture. In recent years, it has emerged as one of Southeast Asia’s most dynamic data centre hubs. Per data centre consultancy DC Byte, more than two-thirds of the region’s new data centre capacity has been committed here.
Based on survey results, our performance is equally impressive. The 2024 Oxford Insights Government AI Readiness Index places Malaysia second in Southeast Asia and 24th globally (out of 188 countries). Yet, despite these achievements, it is undeniable that Malaysia remains largely an adopter, rather than a producer, in the AI value chain.
On the bright side, our country possesses a number of distinctive strengths that can be leveraged to create a differentiated sovereign AI strategy. Robust digital infrastructure and a relatively strong governance framework provide solid foundation for a “small-nation, high-leverage” approach, one that will allow Malaysia to shape the region’s AI trajectory across selected domains, rather than merely consuming technologies developed elsewhere.
Building onto the dimensions of sovereignty
listed above, the following is an exploration of what achieving AI sovereignty could look like for Malaysia.
COMPUTE
Malaysia’s most tangible advantage today lies in digital infrastructure. We sit at the heart of the regional data centre boom, underpinned by abundant and competitively priced land, water, energy and fibre connectivity. Energy, in particular, is a critical differentiator, as it accounts for 60% to 70% of the operational cost for AI compute. Malaysia offers some of the lowest electricity rates in the region.
That said, our sovereign AI ambitions must, in the most emphatic terms, NOT rely solely on expanding data centre capacity. Rapid growth is already straining energy and water resources in digital corridors such as Johor and Greater KL. Even more critically, focusing narrowly on hosting infrastructure for foreign hyperscalers adds limited strategic value to our AI capabilities.
A more sustainable approach requires managing our resources efficiently to balance digital growth while addressing domestic needs. The sovereign AI cloud initiative announced in Budget 2026 is a positive step. Led by the Malaysian Communications and Multimedia Commission (MCMC), the RM2 billion sovereign cloud plan aims to develop government-controlled infrastructure that stands to enhance our decision-making autonomy over AI workloads.
That said, the initiative’s potential will only be realised if it is aligned with broader national objectives. To make the investment worthwhile, it is imperative that the sovereign cloud contributes meaningfully to national capabilities, instead of merely expanding our hardware footprint. That is to say, while proposed as an infrastructure project, the initiative should go beyond mere domestic hosting capacity. Rather, it should be conceived as a foundational enabler for our broader AI ecosystem — supporting AI model development, consolidating public data, monitoring of public services and improving decision-making, fostering workforce readiness (through local participation) and driving innovation, all the while ensuring that sensitive data remains under local jurisdiction.
As for the hardware components like GPUs where technological complexity and economies of scale make full localisation impractical, Malaysia should look to deepen partnerships with global providers and incorporate open-source architecture. The guiding principle being, as outlined in the outset, to align infrastructure development with national needs without overextending our resources.
DATA
While frequently underappreciated, data forms the backbone of any claim to sovereignty. After all, an AI model is only as effective as the data that it has been trained on. Malaysia generates vast quantities of data that reflect our uniquely multicultural and multilingual local context. The challenge is not in volume, but in stewardship: how data is managed, governed and regulated.
Existing frameworks provide a foundation for data governance. While not AI-specific, the Personal Data Protection Act 2010 (PDPA) and sector-specific Risk Management in Technology (RMiT) set standards for data privacy and security, with recent PDPA amendments bringing it closer to the EU’s General Data Protection Regulation (GDPR) gold standard. For government-held data, the Data Sharing Act 2025 enables federal agencies to share datasets more easily, laying the groundwork for cross-agency collaboration.
But beyond tightening regulations, efforts should also focus on how existing data is utilised. Specifically, special care should be taken to strategically harness data in sectors where Malaysia holds a measurable advantage: areas of historical strength, such as plantations and electrical and electronics (E&E); unique sectors such as our halal economy; as well as potential drivers of future growth, including tourism, healthcare and education (more on why we consider these sectors promising pivots for the economy will follow in upcoming articles, so stay tuned).
MODELS
Today, home-grown AI models are emerging.
However, it is important to remember that the strategic value of domestic models is not determined by size or parameter count. As with data, the true worth of sovereign AI models lies in their targeted applicability. Attempting to replicate the full breadth of global AI capabilities is simply too inefficient an endeavour and risks diverting focus from more meaningful objectives.
As it stands, AI models need not be built from scratch. Sovereignty can be achieved through adapting and fine-tuning existing open-source models. The salient point lies in the ability to maintain control over performance, alignment and deployment of the models, thus avoiding black-box dependency on external application programming interfaces, or APIs.
Consequently, open-source frameworks provide a pragmatic balance between innovation and oversight. Many sovereign AI efforts worldwide do in fact appear to be converging around such approaches, including Germany’s SOOFI (Sovereign Open Source Foundation Models) and Switzerland’s Apertus (Latin for “open”). Far from undermining sovereignty, openness allows countries to leverage global innovation while retaining governance controls, a balance that Malaysia would do well to prioritise for its strategic autonomy.
TALENT
Malaysia has long grappled with brain drain, and AI risks intensifying workforce challenges through occupational displacement, skill mismatches and increased competition for specialised talent. The issue is therefore not just one of headcount, but also strategic capability: whether Malaysia can develop, retain and continuously renew the human capital needed for a thriving digital and AI economy.
At the foundational level, MyDIGITAL’s Rakyat Digital platform aims to raise baseline AI literacy by providing free self-learning modules on AI for the public. The next phase, however, must move beyond awareness and basic understanding towards practical proficiency; which is to say, equipping workers with skills to apply AI effectively in industry-relevant contexts.
The formation of MyMahir marks a step in the right direction by centralising skill discovery, career mapping and training pathways, helping address the long-standing fragmentation in Malaysia’s talent ecosystem. Its current value lies in bringing disparate programmes under one platform. The next strategic step is to build on this foundation, moving from mere aggregation towards initiatives that actively channel talent into priority domains aligned with national capacity needs.
Public-private partnerships should similarly evolve from ad hoc training commitments into more durable talent pipelines. Initiatives such as Microsoft’s AI for MY Future, which committed to upskilling 800,000 Malaysians by end-2025, demonstrate the scale that industry collaboration can achieve.
Ultimately, developing local talent must be treated as a long-term strategic investment rather than a short-term mitigation measure. As such, this also includes integrating AI competencies earlier into our formal education system and rethinking incentives to retain high-value talent. Absent of such investments, Malaysia risks underutilising its data, models and infrastructure without the talent that can fully harness them.
REGULATION
The Malaysian government has long recognised AI as a key enabler of the economy, as reflected in formal policy documents such as the National Fourth Industrial Revolution (4IR) Policy and Malaysia Digital Economy Blueprint (MDEB), both published in 2021. Building on this as well as other existing policy frameworks, the establishment of the National Artificial Intelligence Office (NAIO) in December 2024 has strengthened efforts to steer AI development. NAIO acts as the central authority driving national AI strategy, consolidating the policymaking, coordination and implementation of AI development to accelerate its adoption across the economy and society.
Effective regulation, however, requires more than recognition and central coordination. The oversight of AI’s growing capabilities depends on balancing three key objectives: managing its associated risks while simultaneously enabling innovation and ensuring strategic alignment with national priorities in support of sovereignty.
Sector-specific regulatory measures are critical. This is because AI risks vary by industry, necessitating a tailored approach. In healthcare, for example, robust encryption and privacy safeguards are essential for protecting patient records, while financial services require strong fraud detection and credit-scoring systems.
Further, regulatory engagement does not stop at national borders. Regional cooperation is crucial for strengthening Asean’s collective voice on the global stage. The significance of international cooperation in turn underscores the need for us to have a clear understanding of our own AI capabilities. Only by assessing Malaysia’s strengths, weaknesses and forward-looking goals can we then negotiate with clarity on what should remain local and regulated, and what can be shared.
Our weakness in Malaysia is EXECUTION. The inability to focus and empower an aggressive rollout, instead of a pigeonholed approach. To articulate a grand vision so that the people know and understand the purpose and direction, and the opportunities. This is where Singapore excels.
Cross-strait example
In terms of implementation, Singapore offers a clear example.
Rather than pursuing full-stack integration across the AI stack, Singapore’s National AI Strategy (NAIS) 2.0 focuses on deepening partnerships with global compute providers while maintaining a limited — but strategically critical — subset of domestic infrastructure.
A core pillar of Singapore’s strategy is its emphasis on AI solutions tailored to local and regional needs. The city state’s strategy hinges on carving out a distinct niche in AI solutions with localised and highly-specific use cases.
As Laurence Liew, director of Singapore’s national AI and innovation programme (AI Singapore), notes, “[the digital giants] tend to play a horizontal game”, prioritising broad applicability over relevance to local needs. This, in turn, leaves gaps for specialised AI products, especially in areas such as large language models (LLMs) and computer vision, where localisation matters.
Accordingly, a key pillar of Singapore’s strategy lies in nurturing its thriving start-up ecosystem. This is achieved through a suite of initiatives ranging from structured support by government bodies, such as AI Singapore and SGInnovate, to accelerator programmes like A*STAR T-UP (Technology for Enterprise Capability Upgrading), which connect SMEs with industry experts to drive product development. Between 2020 and 2024, Singapore led Southeast Asia in AI start-up funding, attracting as much as US$4.6 billion in investments, and the country is on track to realise a 3.1% increase in gross domestic product from AI-driven gains by 2027.
Notably, Singapore’s position as a global financial and economic hub affords the nation with inherent strengths to stand out as a regional AI pioneer. Specifically, a deep talent pool, strong access to capital and established leadership in sectors at the forefront of AI adoption, including finance and banking. These structural advantages allow Singapore to implement an AI strategy grounded in selectivity according to its strength, not scale.
Nonetheless, the key takeaway is clear: Nations should play to their strengths rather than attempt to compete head-on with global leaders across the full AI value chain. By identifying and leveraging these areas of comparative advantage, countries can work towards a pragmatic form of strategic autonomy and gain control over their own digital destiny. This is the essence of AI sovereignty.
Portfolio performance
The Malaysian Portfolio gained 0.9% for the week ended Jan 28 mirroring the relatively upbeat sentiment for the broader market. Malayan Banking (+6.3%), Kim Loong Resources (+1.3%) and Hong Leong Industries (+0.7%) were the biggest winners while the sole loser was Insas Bhd – Warrants C (-33.3%). Total portfolio returns now stand at 210.3% since inception. This portfolio is outperforming the benchmark FBM KLCI, which is down 4% over the same period, by a long, long way.
The Absolute Returns Portfolio, meanwhile, was up 1.2% last week. The gains lifted total portfolio returns to 46.9% since inception. The biggest gainers were SPDR Gold MiniShares Trust (+11.4%), Alibaba (+6.3%) and Kanzhun (+1.1%). At the other end, Ping An Insurance – H (-8.3%), Berkshire Hathaway (-2.1%) and Ping An Insurance – A (-1.7%) were the top losing stocks for the week.
The AI Portfolio also ended higher, gaining 3% and boosting total portfolio returns to 5.5% since inception. The top three gainers were Datadog (+13.9%), Twilio (+13%) and Alibaba (+6.3%) while RoboSense (-10%), Horizon Robotics (-4%) and Naura Technology (-1.3%) were the three losers last week.
Disclaimer: This is a personal portfolio for information purposes only and does not constitute a recommendation or solicitation or expression of views to influence readers to buy/sell stocks, including the particular stocks mentioned herein. It does not take into account an individual investor’s particular financial situation, investment objectives, investment horizon, risk profile and/or risk preference. Our shareholders, directors and employees may have positions in or may be materially interested in any of the stocks. We may also have or have had dealings with or may provide or have provided content services to the companies mentioned in the reports.
