SG-O-01 | The AI Mega Trend — Singapore's Strategy, Stakes, and Vulnerabilities

Document Code: SG-O-01 Level Designation: Thematic Analysis Version Date: 2026-03-17 Coverage Period: 2017–2026 Primary Sources Consulted:

1. National AI Strategy 1.0 (NAIS 1.0), Smart Nation and Digital Government Office, November 2019
2. National AI Strategy 2.0 (NAIS 2.0), "AI for the Public Good, For Singapore and the World," December 2023
3. Budget 2024 Statement, Prime Minister and Minister for Finance Lawrence Wong, February 2024
4. Budget 2025 Statement, Prime Minister Lawrence Wong, February 2025
5. Budget 2026 Statement, Prime Minister Lawrence Wong, 12 February 2026
6. Minister Josephine Teo, Speech at Singapore Computer Society Tech3 Forum, 22 August 2024
7. Minister Josephine Teo, Committee of Supply 2026 Speech, "Building Singapore's Capability Advantage in a Digital Age," March 2026
8. Minister Josephine Teo, Remarks at World Economic Forum, Davos, 22 January 2026
9. IMF Selected Issues Paper, "Impact of AI on Singapore's Labor Market" (SIP/2024/040), August 2024
10. IMDA, Model AI Governance Framework (2019, 2020), Model AI Governance Framework for Generative AI (2024), Model AI Governance Framework for Agentic AI (2026)
11. AI Verify Foundation, AI Verify Testing Framework and Toolkit, 2022–2025
12. IMDA & AI Verify Foundation, Global AI Assurance Pilot, February 2025
13. MAS, Principles to Promote Fairness, Ethics, Accountability and Transparency (FEAT), 2018
14. MAS, Guidelines for AI Risk Management (Consultation Paper), 2025
15. IMDA, Singapore Digital Economy Report 2025
16. Elections (Integrity of Online Advertising) (Amendment) Act 2024
17. PDPC, Advisory Guidelines on Use of Personal Data in AI Recommendation and Decision Systems, 2024
18. AI Singapore, Annual Reports and Programme Documentation, 2017–2026
19. DBS Group Holdings, Annual Report 2024 and CEO Piyush Gupta public remarks, February 2025
20. Google, Economic Impact Report (Access Partnership), 2024
21. Various speeches by PM Lawrence Wong and Minister Josephine Teo, 2024–2026

Related Documents:

• SG-D-17: Technology, Innovation, and the Smart Nation (1980–2026)
• SG-D-04: Economic Strategy (1959–2026)
• SG-D-02: Education (1959–2026)
• SG-D-03: Defence and National Service (1965–2026)
• SG-D-14: Finance, MAS, and the Financial Centre (1968–2026)
• SG-D-27: POFMA — Design, Application, and Controversy (2019–2026)
• SG-D-10: Labour, Manpower, and the Foreign Worker Question (1960–2026)
• SG-E-25: Singapore's Digital Economy (1998–2026)
• SG-E-26: SkillsFuture (2015–2026)
• SG-E-15: Research, Innovation and Enterprise (2006–2026)
• SG-F-22: Cyber Security as National Strategy (2015–2026)
• SG-F-12: US-China Rivalry and Singapore's Positioning (2017–2026)
• SG-E-36: Crypto, Fintech, and Family Office Regulation (2015–2026)
• SG-E-39: The Gig Economy and Platform Worker Regulation (2015–2026)
• SG-O-04 | Domestic Mega Trends — Singapore's Internal Transformation (2024–2026)
• SG-O-03 | Geopolitical Mega Trends — Singapore in a World on Fire (2024–2026)
• SG-E-01 | The Economic Development Board: Complete Institutional History (1961–2026)
• SG-O-07 | Digital Governance — The GovTech State and Algorithmic Administration

---

1. Key Takeaways

Artificial intelligence is the first technological revolution that Singapore's government has placed at the centre of national strategy from the outset, rather than adapting to belatedly. The National AI Strategy 1.0 (2019) and 2.0 (2023), the formation of a National AI Council chaired by the Prime Minister himself (2026), and the allocation of over S$1 billion in direct AI investment mark a level of state commitment that surpasses even the Smart Nation initiative of 2014. AI is not being treated as a sectoral technology policy. It is being treated as a question of national survival.

Singapore's regulatory approach — voluntary governance frameworks rather than binding legislation — is a deliberate strategic wager. Unlike the European Union, which enacted the AI Act with legally enforceable risk-based classifications, Singapore has chosen to govern AI through a sequence of voluntary Model AI Governance Frameworks (traditional AI in 2019/2020, generative AI in 2024, agentic AI in 2026), supplemented by the AI Verify testing toolkit. This approach — described by the government as "innovation-friendly" — is designed to attract global AI companies and investment. The wager is that market incentives and reputational pressure will achieve compliance without the rigidity of law. If the wager fails — if algorithmic harms accumulate without adequate redress — the absence of binding regulation will be the gap through which injury flows.

The economic opportunity is real, but so is the displacement risk. Singapore has one of the world's highest rates of AI exposure in its labour force. The IMF estimates that 77 percent of Singapore's employed workers are in occupations highly exposed to AI — far above the 60 percent average for advanced economies. Roughly half of these workers are in roles where AI complements their work (managers, engineers, health professionals, lawyers); the other half are in roles where AI threatens to substitute for their labour (clerical workers, administrative professionals, ICT support staff). DBS Bank's announcement in February 2025 that it would eliminate 4,000 contract and temporary positions over three years as AI replaces tasks — with CEO Piyush Gupta stating "for the first time, I'm struggling to create jobs" — is the first major signal of what is coming.

Singapore's position as an AI infrastructure hub creates a new geopolitical vulnerability. Singapore has become NVIDIA's second-largest market after the United States, accounting for roughly one-fifth of NVIDIA's revenue. This reflects genuine demand — from Singtel's GPU-as-a-Service, Google's US$5 billion data centre campus, and the Jurong Island low-carbon data centre park — but also reflects Singapore's use as a trans-shipment point in the US-China chip war. Investigations have revealed Singapore-based intermediaries routing restricted NVIDIA chips to Chinese entities, and Chinese AI companies (including the team behind the Manus AI agent) relocating to Singapore to gain chip access. This places Singapore in a position of strategic sensitivity that its diplomats have long sought to avoid: being pulled into the US-China technology decoupling as a participant rather than a bystander.

---

2. Singapore's AI Strategy — The Arc from 2019 to 2026

2.1 National AI Strategy 1.0 (November 2019)

Singapore's first National AI Strategy was launched in November 2019 by the Smart Nation and Digital Government Office. The strategy was framed as a response to a narrow question: in which domains could AI deliver the greatest social and economic returns for Singapore? The answer was five National AI Projects:

1. Transport and Logistics — Intelligent freight planning to optimise Singapore's role as a shipping hub.
2. Smart Cities and Estates — Seamless and efficient municipal services, including predictive maintenance for public infrastructure.
3. Healthcare — Chronic disease prediction and management, leveraging Singapore's comparatively rich and centralised health data.
4. Safety and Security — Passport-free border clearance via facial recognition and related biometric technology.
5. Education — Adaptive and personalised learning tools.

The strategy also identified five enablers — a "Triple Helix" partnership among the research community, industry, and government; AI talent and education; data architecture; a progressive regulatory environment; and AI research and innovation — and committed S$500 million to AI research, innovation, and enterprise.

The 2019 strategy was cautious in tone. It positioned AI as an opportunity to be seized rather than a force to be reckoned with. It did not address the question of job displacement with any specificity, and its treatment of regulatory risk was limited to a restatement of the Model AI Governance Framework that IMDA had already published in January 2019. The five national projects were sector-specific and bounded — they did not suggest that AI would reshape the entire economy.

2.2 National AI Strategy 2.0 (December 2023)

By December 2023, when then-Deputy Prime Minister Lawrence Wong launched NAIS 2.0 at the Singapore Conference on AI, the framing had changed fundamentally. The title — "AI for the Public Good, For Singapore and the World" — signalled both a broadening of ambition and a recognition that the AI landscape had been transformed by the arrival of large language models, particularly following the release of ChatGPT in November 2022.

NAIS 2.0 introduced three strategic shifts:

1. From opportunity to necessity. AI was repositioned from an economic opportunity to a national imperative. The language shifted from "seize the moment" to "AI is essential for Singapore's continued relevance and competitiveness."

2. From local to global. The 2019 strategy had focused on domestic applications. NAIS 2.0 explicitly framed Singapore as a global node in the AI ecosystem — a contributor to AI governance norms, a host for AI infrastructure, and a bridge between Western and Asian AI development.

3. From projects to infrastructure. The five discrete national projects were replaced by a systemic approach. NAIS 2.0 identified three interconnected systems — Activity Drivers (industry, government, research), People and Communities (talent, capabilities, placemaking), and Infrastructure (compute, data, trusted environment, thought leadership) — supported by ten enablers and fifteen key actions.

The strategy set two overarching goals: Excellence (developing "peaks of excellence" in AI to maximise value creation) and Empowerment (enabling individuals, businesses, and communities to use AI with confidence and trust). The dual framing was significant: it acknowledged that AI's benefits would not flow automatically to all segments of society, and that active policy intervention would be required to ensure broad-based participation.

2.3 Budget 2024: The Billion-Dollar Commitment

Then-Deputy Prime Minister and Minister for Finance Lawrence Wong's Budget 2024 speech on 16 February 2024 translated the strategy into money. He announced: "To support this strategy and further catalyse AI activities, I will invest more than $1 billion over the next five years into AI compute, talent, and industry development." This was among the largest single commitments to a specific technology domain in Singapore's fiscal history.

The investment was allocated across three pillars:

• Compute infrastructure — Expanding the availability of GPU capacity for AI training and inference, including public-sector compute resources.
• Talent development — Scaling AI Singapore's apprenticeship programme, expanding the TechSkills Accelerator (TeSA), and attracting global AI talent.
• Industry development — Supporting enterprise AI adoption through grants, sandboxes, and the Productivity Solutions Grant (PSG).

2.4 Budget 2026: The National AI Council and AI Missions

The February 2026 Budget marked the culmination of this strategic arc. Prime Minister Wong devoted an entire section of his Budget speech to AI under the heading "Harness AI as a Strategic Advantage," declaring:

"In a changed world, a decisive factor for success will be how we harness new technologies — foremost among them, artificial intelligence."

"Harnessed well, AI will be a strategic advantage for Singapore. It can help us overcome our structural constraints — our limited natural resources, rapidly ageing population and tight labour market."

"Our advantage does not lie in building the latest frontier models. It lies in deploying AI effectively, responsibly, and at speed."

The key institutional innovation was the formation of a National AI Council, an inter-ministerial body chaired by PM Wong himself, with Deputy Prime Minister Gan Kim Yong and Health Minister Ong Ye Kung among its members. The council was tasked with providing strategic direction for Singapore's AI agenda, commissioning "AI Missions" in priority sectors, and — critically — "unlocking regulations and resources" to accelerate AI deployment. The four initial mission areas were advanced manufacturing, connectivity and logistics, finance, and healthcare.

The Budget also announced:

• Enterprise Compute Initiative — S$150 million to provide cloud credits, AI tools, and consultancy services to help Singapore-based companies develop AI minimum viable products (MVPs).
• Enterprise Innovation Scheme expansion — AI expenditures qualifying for tax deduction, capped at S$50,000 per year of assessment for 2027–2028.
• S$3 billion top-up to the National Productivity Fund — Enabling companies to tap AI and other high-value technology.
• Free premium AI tools — Six months of free access to premium AI tools for Singaporeans who complete selected AI courses.
• SkillsFuture redesign — Clearer AI learning pathways on the SkillsFuture website.

PM Wong also gave a pointed shout-out to DBS and Grab as examples of Singaporean companies leading in AI adoption — a notable choice given that DBS had just announced the elimination of 4,000 jobs due to AI, and Grab had opened an AI Centre of Excellence in Singapore.

---

3. Key Policies and Institutions

3.1 AI Singapore (AISG)

AI Singapore was established in 2017 as a national programme under the National Research Foundation, with funding of S$150 million (later expanded). Its mandate spans the entire AI development chain: fundamental research, applied R&D, talent development, and industry adoption.

AISG's most significant contributions are:

The AI Apprenticeship Programme (AIAP). Launched in 2018, AIAP is a nine-month, full-time, fully subsidised programme that trains Singaporeans in AI engineering through two months of deep skilling followed by seven months of on-the-job training on real-world industry projects. Apprentices receive a monthly stipend of S$3,500 to S$5,500. The programme reports a graduate placement rate exceeding 90 percent, with alumni securing roles as AI engineers, MLOps engineers, and data scientists. By 2026, AIAP has expanded to include AIAP for Industry (a six-month intensive track), the AI Internship Programme (AIIP), and the LLM Application Developer Programme (LADP). Batch 22 opens for applications on 26 January 2026.

SEA-LION (Southeast Asian Languages in One Network). This is arguably AISG's most strategically significant project. SEA-LION is a family of open-source, multilingual, multimodal large language models designed specifically for Southeast Asian languages, cultures, and contexts. Trained on eleven languages — including Indonesian, Thai, Vietnamese, Filipino, Burmese, Malay, Lao, Khmer, and Tamil — SEA-LION addresses the fundamental problem that the world's dominant LLMs are trained overwhelmingly on English-language data and reflect Western cultural assumptions. SEA-LION is part of the National Multimodal LLM Programme (NMLP), which received S$70 million in funding from the National Research Foundation and also supports A*STAR's MERaLiON model.

The strategic logic is sovereign AI: Singapore's position is that reliance on foreign-built foundation models creates a dependency that is incompatible with national sovereignty. As IMDA stated in December 2023, there is a "strategic need to develop sovereign capabilities in LLMs." SEA-LION is Singapore's answer — not a frontier model competing with GPT or Gemini, but a regionally attuned model that serves populations whose languages are under-represented in global AI training data. GoTo's Sahabat-AI ecosystem, built on SEA-LION for Indonesian developers and integrated into the Gojek and GoPay platforms, is the most prominent commercial deployment.

100 Experiments. AISG's "100 Experiments" programme matches companies with AISG researchers to co-develop AI solutions for specific business problems, with AISG absorbing 70 percent of the project cost. The programme has partnered with companies across manufacturing, logistics, healthcare, and financial services.

3.2 The Government Technology Agency (GovTech)

GovTech, the government agency responsible for the public sector's technology infrastructure, has become the primary vehicle for AI deployment within the Singapore government. Its AI work falls into three categories:

Pair Suite. Launched by Open Government Products (OGP), an experimental technology team within GovTech, the Pair suite is the most widely adopted AI toolset in the Singapore public service. Pair Chat, the flagship product, is a government-controlled chatbot powered by the same large language models as ChatGPT but deployed within a secure government environment. It is available on all government-issued devices.

Adoption has been rapid: about 80 percent of 150,000 public officers have used Pair Chat, with over 20,000 weekly active users. Beyond Pair Chat, the suite includes:

• Pair Noms — An LLM-powered tool for transcribing, formatting, and generating meeting minutes. Pilot users report saving approximately 50 percent of the time typically spent on meeting minutes — a task that can take civil servants anywhere from a few hours to a full working day.
• AIBots — A platform enabling officers to create custom chatbots for specific tasks (HR queries, budget guidance, procurement processes). Over 20,000 such bots have been created across government agencies.

LaunchPad. Built on Microsoft Azure OpenAI Service, LaunchPad is an internal government platform where agencies can explore AI prototypes, browse central products, and learn from each other's experiments. With more than 3,000 monthly active users across all government agencies, LaunchPad has generated over 500 ideas and 20 prototypes. It operates through ideathons and structured sprints, targeting transformation teams and public officers across a six-month cycle.

Transcribe. GovTech's speech-to-text platform, capable of recognising localised Singaporean English (Singlish) and generating transcripts of interviews, speeches, and meetings.

VICA (Virtual Intelligent Chat Assistant). The platform underpinning government chatbots on public-facing websites, replacing the earlier "Ask Jamie" chatbot that had been deployed across more than 70 government websites.

3.3 The Infocomm Media Development Authority (IMDA)

IMDA's role in the AI ecosystem is regulatory and convening. It is the institutional author of Singapore's AI governance frameworks, the operator of the AI Verify testing toolkit, and the coordinator of the data centre expansion programme. Its key AI-related functions include:

• Publishing and maintaining the Model AI Governance Framework series (traditional AI, generative AI, agentic AI).
• Operating the AI Verify Foundation.
• Managing the Data Centre Call for Applications process.
• Co-leading Digital Industry Singapore (DISG) with EDB.
• Administering the Productivity Solutions Grant for AI-enabled solutions.

3.4 Digital Industry Singapore (DISG)

DISG is a joint initiative of IMDA and the Economic Development Board (EDB). It focuses on attracting and growing digital economy companies in Singapore, with AI as the primary growth vector. DISG supported Grab's AI Centre of Excellence and has facilitated investments by Google, Microsoft, Amazon Web Services, and NVIDIA in Singapore's AI infrastructure.

3.5 The Monetary Authority of Singapore (MAS) — FEAT and Project Veritas

MAS has developed sector-specific AI governance for the financial industry through the FEAT principles (Fairness, Ethics, Accountability, Transparency), published in 2018. The principles were operationalised through Project Veritas, which provides financial institutions with a toolkit for evaluating their AI and data analytics solutions against the FEAT principles. The Veritas Toolkit 2.0 was released in 2023.

In 2025, MAS issued a consultation paper proposing formal Guidelines on AI Risk Management for financial institutions — a step toward more prescriptive governance in the sector most heavily adopting AI. The guidelines cover model risk management, data governance, third-party AI, and consumer protection.

---

4. Economic Impact — What AI Means for the Singapore Economy

4.1 The Digital Economy Reaches 18.6 Percent of GDP

According to IMDA's Singapore Digital Economy Report 2025, Singapore's digital economy expanded by S$12 billion in 2024 to a total of S$128.1 billion — 18.6 percent of GDP, up from 18.0 percent in 2023 and 14.9 percent in 2019. The growth was driven by widespread digitalisation and, increasingly, AI adoption across all sectors.

The most striking data point: SME AI adoption more than tripled in a single year, rising from 4.2 percent in 2023 to 14.5 percent in 2024. Among non-SME firms, AI adoption jumped from 44 percent to 62.5 percent over the same period. AI-adopting SMEs deployed AI across an average of three business functions, while non-SMEs used AI across five. The most common applications were in IT, customer service, and finance/accounting.

SMEs using AI-enabled solutions under the Productivity Solutions Grant achieved average cost savings of 52 percent; those adopting AI-powered cybersecurity solutions achieved savings of 71 percent.

Singapore's technology workforce expanded from 208,300 in 2023 to 214,000 in 2024, with AI and data roles and cybersecurity roles growing fastest. Significantly, more than two-thirds of this digital growth occurred in non-technology sectors — manufacturing, financial services, logistics, and professional services — reflecting the permeation of AI across the broader economy.

4.2 Projected Economic Benefits

A Google-commissioned Economic Impact Report by Access Partnership (2024) estimated that Singapore businesses could gain US$147.6 billion in economic benefits — measured in cost savings, revenue increments, and productivity gains — by 2030, if AI-powered products and solutions are fully utilised. While commissioned research from technology companies warrants scrutiny for optimism bias, the order of magnitude is consistent with broader estimates: McKinsey Global Institute projects that generative AI alone could add the equivalent of US$2.6 to US$4.4 trillion annually to the global economy, and PwC estimates AI could contribute up to 14 percent of global GDP by 2030.

For Singapore specifically, the AI investment boom is expected to sustain growth through 2026. Electronics and ICT sectors — driven by AI-related semiconductor and server demand — were key contributors to Singapore's 5 percent GDP growth in 2025, which exceeded consensus estimates.

4.3 The Infrastructure Investment Surge

The scale of AI-related infrastructure investment flowing into Singapore is without precedent for a city-state of 6 million people:

• Google — Total investments in Singapore technical infrastructure reached US$5 billion (up from US$850 million in 2018), with a fourth data centre facility completed in 2024 at the Jurong West campus. More than 500 people work in Google's Singapore data centres.

• NVIDIA and Singtel — NVIDIA partnered with Singtel in early 2024 to launch "AI-ready" data centre infrastructure and GPU-cloud services for Southeast Asia, initially powered by NVIDIA H100 Tensor Core GPU clusters, with plans to deploy GB200 Grace Blackwell Superchips. The SIT x NVIDIA AI Centre (SNAIC), launched at Singapore Institute of Technology's Punggol campus, has initiated more than 50 projects and forged over 20 industry partnerships. It will train more than 200 practitioners over three years.

• Data Centre Expansion — Singapore imposed a moratorium on new data centre development in 2019 to manage energy consumption and emissions. The moratorium was lifted in 2022, with the first Data Centre Call for Applications (DC-CFA) targeting approximately 60 MW of new capacity under a selective approval regime. In 2024, IMDA announced a target of at least 300 MW of additional capacity. In October 2025, approximately 20 hectares of land on Jurong Island was set aside for Singapore's largest low-carbon data centre park, with the potential to accommodate up to 700 MW of capacity.

• Broader Regional Investment — Singapore has attracted an estimated US$26 billion in technology infrastructure investment from global tech companies. Meta announced funding for hyperscale campus development in Singapore. AWS expanded data centre infrastructure across the region.

The tension between data centre expansion and Singapore's climate commitments — the country's Green Plan 2030 and its net-zero by 2050 target — is a running policy challenge. The Green Data Centre Roadmap (2024) imposes stricter energy efficiency, renewable energy sourcing, and cooling-system innovation requirements on new facilities, but the sheer scale of AI compute demand creates an inherent conflict with emissions reduction goals (see SG-D-25).

---

5. The AI Talent Question

5.1 The Talent Gap

Singapore aims to triple its AI practitioner pool from 4,500 to 15,000 by 2029. This target, set in the context of NAIS 2.0, reflects both the scale of demand and the severity of the current shortage. Survey data shows that 81 percent of employers prioritise AI talent hiring, but 74 percent struggle to find qualified candidates.

The talent pipeline draws from multiple sources:

• AIAP and AISG programmes — The flagship apprenticeship produces approximately 50 to 60 graduates per batch. The expanded portfolio (AIAP for Industry, AIIP, LADP) is designed to scale output, but the numbers remain small relative to demand.

• University programmes — NUS, NTU, SMU, and SUTD all offer AI-related degree programmes, but the university pipeline produces generalists who require further specialisation.

• TechSkills Accelerator (TeSA) — Has placed and trained 17,000 locals in AI and tech roles since its launch in 2016.

• SG Digital Scholarship — S$20 million fund for AI-related education and overseas internships over three years.

• Industry training — AWS committed to training 5,000 individuals annually from 2024 to 2026 through its AI Spring programme. Microsoft's Asia AI Odyssey targets 30,000 developers across ASEAN. Google's training programmes reach thousands of Singaporeans through partnerships with local institutions.

5.2 Foreign Talent and Global AI Attraction

Singapore's AI talent strategy relies heavily on attracting foreign expertise — a continuation of the open immigration model that has defined Singapore's economic strategy since independence (see SG-D-10, SG-G-29). The tensions that have surrounded foreign talent policy in other sectors — construction, domestic work, professional services — are beginning to surface in AI as well.

The government's position is that Singapore must be a magnet for the world's best AI researchers and engineers, and that their presence creates spillover benefits for local workers and institutions. Critics argue that heavy reliance on foreign AI talent risks creating an AI workforce that is internationally mobile and not anchored to Singapore's long-term interests.

The establishment of the SIT x NVIDIA AI Centre (SNAIC) represents an attempt to bridge this gap — a facility that combines international expertise (NVIDIA's technology and research capabilities) with local institutional anchoring (SIT's student body and industry partnerships). SNAIC's partnerships with Monash University in Australia, Chulalongkorn University in Thailand, and VNU-UEL in Vietnam extend its reach regionally, positioning Singapore as a talent development hub for the broader Southeast Asian AI ecosystem.

---

6. AI in Government — The Internal Revolution

6.1 The Scale of Adoption

The Singapore government's internal adoption of AI is among the most advanced of any government in the world. The statistics are striking:

• 80 percent of 150,000 public officers have used Pair Chat.
• Over 20,000 custom AIBots created by public officers for specific tasks.
• Over 3,000 monthly active users of LaunchPad across all government agencies.
• Over 500 ideas generated through LaunchPad ideathons.
• 50 percent time savings reported by pilot users of Pair Noms for meeting minutes.

These are not pilot numbers. They represent a systemic shift in how the Singapore civil service operates. The Pair suite has been described by UNDP's Policy Centre in Singapore as one of the most significant examples of AI deployment in government anywhere in the world.

6.2 The Policy Logic

The government's rationale for aggressive internal AI adoption is threefold:

1. Productivity. Singapore's civil service faces the same labour constraints as the private sector — an ageing population, tight labour market, and rising expectations from citizens. AI enables the civil service to do more with the same or fewer staff.

2. Credibility. A government that deploys AI successfully within its own operations is better positioned to advise the private sector on AI adoption. The government's ability to speak from experience — rather than theory — enhances the credibility of its AI governance frameworks.

3. Control. By deploying government-built AI tools (rather than relying on commercial platforms), the government maintains control over data flows, security protocols, and the terms of AI interaction. Pair Chat runs on government infrastructure, not on OpenAI's servers.

6.3 Risks of Government AI Adoption

The government's internal AI deployment raises several risks that have received less public attention:

• Automation bias. Public officers who rely on AI-generated drafts, summaries, and recommendations may develop an over-reliance on AI outputs, reducing the quality of independent judgment. The risk is particularly acute in policy analysis, where AI's tendency to pattern-match from existing documents can reinforce conventional thinking at the expense of creative or dissenting analysis.

• Centralisation of information flows. If AI tools are used to summarise ministerial briefings, draft parliamentary questions, and generate policy options, the AI becomes a filter through which information flows to decision-makers. The biases embedded in the AI — including training data biases and prompt design choices — become invisible but consequential influences on government decision-making.

• Data security. The Pair suite operates within the government's secure environment, but the expansion of AI tools across 150,000 public officers increases the surface area for data exposure. The risk is not a dramatic breach but a gradual erosion of information boundaries — an officer inadvertently feeding sensitive information into an AI tool that stores, indexes, or surfaces it inappropriately.

---

7. Singapore's Regulatory Approach — The Innovation-First Wager

7.1 The Model AI Governance Framework Series

Singapore's AI governance architecture is built on a series of voluntary, principles-based frameworks rather than binding legislation. This is a deliberate choice, and the government has been explicit about the reasoning: binding regulation risks deterring investment and slowing innovation, while voluntary frameworks can evolve rapidly with the technology.

The framework has evolved through four iterations:

1. Model AI Governance Framework, First Edition (January 2019) — Launched at the World Economic Forum in Davos, this was one of the world's first government-published AI governance frameworks. It set out principles for transparency, fairness, and accountability in AI deployment, with a focus on traditional machine learning systems.

2. Model AI Governance Framework, Second Edition (January 2020) — Updated with implementation guidance, case studies, and self-assessment tools.

3. Model AI Governance Framework for Generative AI (May 2024) — Developed by the AI Verify Foundation and IMDA, this extended the framework to address the specific challenges of generative AI: hallucination, content provenance, model training on personal data, and the blurring of human and machine-generated content.

4. Model AI Governance Framework for Agentic AI (January 2026) — Announced by Minister Josephine Teo at the World Economic Forum in Davos on 22 January 2026, this was described as the first-of-its-kind governance framework for agentic AI — AI systems that can autonomously take actions, make decisions, and interact with other systems and humans. The framework addresses four dimensions: bounding risk by limiting agents' powers, maintaining meaningful human accountability through significant checkpoints, implementing technical controls, and enabling end-user responsibility through transparency. The framework applies to all organisations deploying agentic AI in Singapore, but compliance remains voluntary.

7.2 AI Verify — The Testing Toolkit

AI Verify is Singapore's most distinctive contribution to the global AI governance landscape. It is a testing framework and open-source software toolkit that enables organisations to conduct voluntary self-assessments of their AI systems against eleven AI ethics principles drawn from internationally recognised frameworks (EU, OECD, Singapore's own governance framework).

The toolkit combines technical tests (bias detection, explainability metrics, robustness testing) with process-based checks (data governance, stakeholder engagement, documentation). By using AI Verify, organisations can generate a standardised report demonstrating how their AI system performs against each principle.

The AI Verify Foundation was established in June 2023 in collaboration with major technology companies — including Google, IBM, Microsoft, Red Hat, and Salesforce — to maintain and develop the toolkit as an open-source community project. By 2025, the Foundation had expanded to over 90 member organisations.

In February 2025, IMDA and the AI Verify Foundation launched the Global AI Assurance Pilot, designed to codify emerging norms and best practices around technical testing of generative AI applications. IMDA also signed a Memorandum of Intent with MLCommons, a leading AI engineering consortium recognised by the US National Institute of Standards and Technology (NIST).

7.3 Contrast with the EU AI Act

The contrast between Singapore's approach and the European Union's AI Act (which entered into force in August 2024) is instructive:

Singapore's government has been careful not to criticise the EU approach directly, but the strategic positioning is unmistakable. As Minister Josephine Teo stated at the SCS Tech3 Forum in August 2024:

"What we thought was meaningful for us to do as a nation was to consider how AI could be used for the public good."

The implicit message: Singapore governs for AI's benefits, not its risks. Whether this framing will survive the first major AI-caused harm in Singapore remains to be seen.

7.4 PDPA and AI — The Data Privacy Dimension

The Personal Data Protection Act (PDPA) provides the legal framework for data privacy in Singapore, but it was not designed for the AI era. In 2024, the Personal Data Protection Commission (PDPC) issued Advisory Guidelines on the Use of Personal Data in AI Recommendation and Decision Systems, clarifying how it interprets the PDPA's requirements in the context of AI. The guidelines address consent, purpose limitation, data minimisation, and the use of personal data for AI training.

However, the guidelines do not address generative AI specifically. The PDPC has acknowledged that generative AI raises "distinct privacy concerns regarding the use of personal data to train foundation models" and stated that these concerns are "currently being studied." The gap between the pace of generative AI deployment and the pace of regulatory guidance creates a window during which organisations operate without clear rules.

7.5 Deepfakes and Elections

Singapore addressed the specific risk of AI-generated deepfakes in elections through the Elections (Integrity of Online Advertising) (Amendment) Bill, passed by Parliament on 15 October 2024. The law prohibits the publication, sharing, boosting, or reposting of deepfake content depicting election candidates — content made using AI or non-AI techniques (Photoshop, dubbing, splicing) that realistically depicts a candidate saying or doing something they did not.

The prohibition applies from the issuance of the writ of election until the close of polling. Penalties include fines of up to S$1,000 and/or imprisonment for up to 12 months for individuals, and fines of up to S$1 million for social media platforms that fail to comply with corrective directions. The law came into effect for the general election held on 3 May 2025.

The law addresses election-specific deepfakes but does not create a general prohibition on deepfakes or AI-generated disinformation. For non-election contexts, the government relies on POFMA (see SG-D-27) — a law designed for a pre-AI era and which requires a minister to identify specific false statements of fact. Whether POFMA's architecture can scale to address the volume and sophistication of AI-generated falsehoods is an open and pressing question.

---

8. Risks and Concerns

8.1 Job Displacement — The IMF Assessment

The most rigorous analysis of AI's impact on Singapore's labour market is the IMF's Selected Issues Paper published in August 2024. Its central finding is stark: 77 percent of Singapore's employed workers are in occupations highly exposed to AI — the highest rate of any country studied and well above the 60 percent average for advanced economies.

The IMF divides this exposed workforce into two groups of roughly equal size:

• 38.9 percent are in occupations with high AI complementarity — managers, science and engineering professionals, health professionals, legal professionals, and teaching professionals. For these workers, AI is a productivity enhancer. Provided they have the skills and infrastructure to use AI effectively, they stand to gain.

• 38.6 percent are in occupations with low AI complementarity — business and administration professionals, ICT professionals and associate professionals, clerical support workers, and some services and sales workers. For these workers, AI is a substitution threat. Their tasks — data entry, report generation, scheduling, routine analysis, customer service — are precisely the tasks that current AI systems perform well.

The IMF estimates that women and younger workers are more exposed to AI's effects, which, absent appropriate policies, could worsen income inequality.

8.2 DBS — The Canary in the Coal Mine

The DBS announcement of February 2025 is the most significant concrete example of AI-driven job displacement in Singapore to date. Key details:

• DBS plans to cut approximately 4,000 contract and temporary staff over the next three years — roughly 10 percent of its total workforce.
• No permanent employees will lose their jobs; the cuts will come from "natural attrition as temporary and contract roles roll off."
• DBS had deployed approximately 800 AI models across 350 use cases by end-2024.
• The measured economic impact of these AI deployments was expected to exceed S$1 billion before end-2025.
• CEO Piyush Gupta stated: "In my 15 years of being a CEO, for the first time, I'm struggling to create jobs. I'm struggling to say how I will repurpose people to create jobs."
• DBS will add approximately 1,000 new positions in AI — a net loss of 3,000 jobs.

The DBS case crystallises the challenge. The bank is not failing; it is thriving. AI is making it more productive and profitable. The S$1 billion in economic impact flows to shareholders, to the remaining workforce through higher-value work, and to customers through better services. But 4,000 workers — disproportionately in contract and temporary roles, disproportionately in lower-skilled functions — are displaced. The government's response — SkillsFuture, AI retraining programmes, the Progressive Wage Model — is designed to catch these workers. Whether it can do so at the pace and scale required is the central test of Singapore's AI social contract.

8.3 AI Bias

Singapore's population is multiracial, multilingual, and multicultural (see SG-D-09, SG-G-01). AI systems trained on global data — overwhelmingly English-language, Western-centric — risk embedding biases that are poorly calibrated to Singapore's context. Specific concerns include:

• Hiring algorithms that disadvantage candidates whose names, educational backgrounds, or communication styles do not match patterns in Western-dominated training data.
• Credit scoring models that systematically disadvantage particular demographic groups.
• Language models that perform poorly in Singlish, Malay, Tamil, or Mandarin — or that embed stereotypes about speakers of these languages.

The AISG-organised AI Safety Red Teaming Challenge (November 2024), which brought over 350 participants from nine countries to test four LLMs for cultural bias stereotypes in non-English languages, represents an initial effort to address these risks. But the scale of the problem — every AI system deployed in Singapore potentially carrying biases from its training data — far exceeds the scale of current testing efforts.

8.4 Data Privacy

The use of personal data to train AI models is the most significant unresolved tension in Singapore's data governance framework. The PDPA's consent-based model — where organisations must obtain consent for the collection, use, and disclosure of personal data — sits awkwardly alongside the reality that foundation models are trained on vast corpora of internet data, much of which includes personal data collected without specific consent for AI training purposes.

The PDPC's Advisory Guidelines (2024) address the use of personal data in AI recommendation and decision systems but explicitly do not address the use of personal data to train foundation models. This gap leaves organisations in a grey zone: they can deploy AI systems trained on potentially nonconsensual data while complying with the PDPA's requirements for their own data handling.

8.5 Concentration of Power

Singapore's AI strategy concentrates significant power in a small number of institutions: the National AI Council (chaired by the Prime Minister), IMDA, GovTech, AI Singapore, and MAS. The governance frameworks are written by the government. The testing toolkit is maintained by a government-linked foundation. The talent pipeline is government-funded. The compute infrastructure is government-directed.

This concentration is characteristic of Singapore's governance model (see SG-D-07) and has obvious efficiency advantages: it enables rapid decision-making, coordinated investment, and coherent strategy. But it also means that critical decisions about AI's deployment — which sectors, which use cases, which risks to tolerate — are made by a small group of officials and technocrats without the public deliberation, independent oversight, or adversarial scrutiny that characterise AI governance debates in other democracies.

---

9. AI and Singapore's Competitive Position

9.1 Financial Hub

Singapore leads the Asia-Pacific region in AI deployment within financial services. According to industry surveys, 62 percent of Singapore financial institutions deployed advanced fraud detection and transaction monitoring in 2025 — the highest rate globally. MAS's FEAT principles and Veritas toolkit give Singapore's financial AI governance a degree of structure and credibility that other Asian financial centres lack.

AI is reshaping the financial sector in several dimensions:

• Compliance and KYC — AI automates know-your-customer checks, anti-money laundering screening, and regulatory reporting. These are labour-intensive tasks that employed thousands of compliance professionals in Singapore's banking sector.
• Wealth management — AI-powered advisory tools enable hyper-personalised portfolio management at scale, potentially democratising services previously available only to high-net-worth clients.
• Insurance — Claims processing, underwriting, and risk assessment are being automated, with DBS's experience providing a preview of the workforce implications.
• Trading — Algorithmic and AI-driven trading strategies dominate high-frequency and quantitative trading, with Singapore positioning itself as an Asia-Pacific hub for these activities.

The risk is that AI enables the financial sector to grow its output without proportionally growing its workforce — the "jobless growth" scenario that PM Wong explicitly addressed in February 2026, stating that Singapore would avoid it.

9.2 Shipping and Logistics Hub

Singapore's position as the world's busiest container trans-shipment hub (see SG-E-08, SG-E-35) is being reshaped by AI. The National AI Mission for connectivity and logistics is designed to accelerate AI adoption in port operations, freight planning, and supply chain optimisation. AI applications include:

• Automated container yard planning and vessel scheduling.
• Predictive maintenance for port equipment.
• Digital twins for port operations simulation.
• AI-optimised shipping route planning.

The Tuas Mega Port (see SG-E-35), scheduled for full operation by 2040, is being designed as a "smart port" with AI integrated into its core operations from the outset.

9.3 Legal Services Hub

Singapore's ambitions as an international legal and dispute resolution hub (see SG-D-08) intersect with AI in complex ways. AI's ability to automate contract review, due diligence, legal research, and document drafting threatens the labour-intensive model on which much of the legal profession is built. The IMF classifies legal professionals as having high AI complementarity — meaning AI enhances rather than replaces their work — but this is true primarily for senior lawyers engaged in judgment-intensive work. Junior lawyers, paralegals, and legal secretaries face significant substitution risk. The IMF notes legal secretaries face 75 percent AI exposure.

9.4 The Geopolitical Dimension — NVIDIA, Chips, and the US-China Squeeze

Perhaps the most consequential and least discussed dimension of Singapore's AI positioning is its emergence as a node in the US-China semiconductor competition.

Singapore accounts for approximately one-fifth of NVIDIA's global revenue — a remarkable figure for a country of 6 million people. This partly reflects genuine domestic and regional demand for GPU compute. But it also reflects Singapore's role as a trans-shipment hub for AI chips destined for markets subject to US export controls.

Recent developments have exposed the risks:

• A Singapore-based company was alleged to have facilitated the transfer of approximately US$2 billion worth of NVIDIA AI processors to Chinese entities, circumventing US export controls. The company allegedly purchased chips under legitimate pretenses and rerouted them through third-party distributors and shell companies.

• The Chinese AI company behind Manus, a general-purpose AI agent, relocated its headquarters from China to Singapore — widely interpreted as a move to facilitate easier access to NVIDIA chips.

• NVIDIA's largest Southeast Asian customer has been linked to potential export control violations, exposing the limits of US enforcement mechanisms.

These developments create a dilemma for Singapore. Being a major AI infrastructure market generates economic benefits: investment, jobs, tax revenue, and strategic relevance. But being perceived as a conduit for circumventing US technology controls risks damaging Singapore's relationship with its most important security partner (see SG-F-02). Singapore's foreign policy has long been premised on maintaining equidistance between the US and China (see SG-F-12). The AI chip issue threatens to force a choice that Singapore's diplomats have spent decades trying to avoid.

---

10. Key Speeches and Positions

10.1 Prime Minister Lawrence Wong

PM Wong has made AI a defining theme of his premiership. Key statements:

Budget 2024 (February 2024):

"To support this strategy and further catalyse AI activities, I will invest more than $1 billion over the next five years into AI compute, talent, and industry development."

Budget 2026 (12 February 2026):

"In a changed world, a decisive factor for success will be how we harness new technologies — foremost among them, artificial intelligence."

"Harnessed well, AI will be a strategic advantage for Singapore. It can help us overcome our structural constraints — our limited natural resources, rapidly ageing population and tight labour market."

"Our advantage does not lie in building the latest frontier models. It lies in deploying AI effectively, responsibly, and at speed."

"Singapore will not be passive in the face of rapid changes around us. We will adapt. We will compete. We will continue to move forward with confidence. By harnessing AI as a strategic advantage, we will shape our own future and secure our place in this changed world."

Budget 2026 Round-Up Speech (26 February 2026): Wong stated that Singapore would avoid "jobless growth" as AI reshapes the global economy. The government would harness AI to expand the economy while ensuring that growth produces good jobs and higher wages.

10.2 Minister Josephine Teo

As Minister for Digital Development and Information, Josephine Teo is Singapore's principal voice on AI policy. She has been prolific in articulating the government's position:

SCS Tech3 Forum (22 August 2024):

"Rather than 'Humans vs AI', it is going to be 'Humans with AI' vs 'Humans without AI'."

"Many thoughtful observers have pointed out that it is not so much AI displacing workers, but AI-proficient workers displacing AI-deficient workers. This is the way my team and I think of AI."

On AI democratisation (2026):

"We want to take full advantage of AI's ability to be democratised, or to put it more simply, for its benefits to spread widely because solutions, once too expensive or complex, are more accessible."

On Singapore's global position:

"International counterparts recognise Singapore's ability to respond holistically across industries, enterprises and the workforce through a range of enablers — from R&D and infrastructure to safety and governance. On the global stage, Singapore is frequently at the table, and its progressive, thoughtful approach to AI makes it a credible partner and useful reference point."

Committee of Supply 2026 (March 2026): Teo announced the National AI Impact Programme, targeting 10,000 enterprises and 100,000 "AI Bilingual" workers — workers who combine domain expertise with AI proficiency. She also announced the Model AI Governance Framework for Agentic AI, enhanced cybersecurity measures for critical infrastructure, and strengthened public service media as a "frontline defence against the spread of AI-driven misinformation."

Latent Space Podcast: In a notable appearance on the Latent Space podcast (a popular AI industry podcast), Teo outlined her vision of "Building the AI Engineer Nation," framing Singapore's ambition not as producing frontier AI researchers but as creating a nation of skilled AI deployers — professionals across every domain who can effectively use AI tools.

10.3 International Positioning

Singapore has positioned itself as an active participant in the emerging global AI governance architecture:

• Bletchley Park AI Safety Summit (November 2023) — Singapore was among the 28 countries that signed the Bletchley Declaration on AI safety.
• AI Seoul Summit (May 2024) — Singapore endorsed the Seoul Statement of Intent toward International Cooperation on AI Safety Science, alongside G7 countries, Australia, and the Republic of Korea.
• AI Action Summit, Paris (February 2025) — Singapore endorsed the Leaders' Statement on Inclusive and Sustainable AI. IMDA and the AI Verify Foundation published the AI Safety Red Teaming Challenge Evaluation Report, based on findings from a challenge event that brought over 350 participants from nine countries to red-team four LLMs for cultural bias in non-English languages.
• AI Safety Institute Network — Singapore is a founding member of the international network of AI Safety Institutes, alongside the US, UK, EU, Japan, South Korea, Canada, France, Kenya, and Australia.

---

11. Connection to the Existing Corpus

AI is not a standalone policy domain. It intersects with virtually every dimension of Singapore governance documented in this corpus:

Smart Nation (SG-D-17)

The National AI Strategy is the latest and most significant iteration of the Smart Nation initiative launched by Lee Hsien Loong in November 2014. Where Smart Nation 1.0 focused on digital government services, sensors, and data infrastructure, the AI phase represents a qualitative leap — from digitalisation (making existing processes digital) to transformation (using AI to do things that were previously impossible). The Smart Nation 2.0 plan includes a S$120 million fund for AI adoption.

Education (SG-D-02, SG-G-15, SG-G-16, SG-G-17, SG-G-18)

The AI talent question intersects directly with Singapore's education system. The NAIS 1.0 identified education as one of five national AI projects. The AI Apprenticeship Programme and SkillsFuture AI courses represent the vocational response. But the deeper question — whether Singapore's education system, with its emphasis on standardised testing and credentialism (see SG-G-16), produces the kind of creative, adaptable thinkers who can thrive in an AI-augmented economy — remains unresolved. The government's answer — "AI Bilingual" workers who combine domain expertise with AI skills — is a workforce strategy, not an education reform.

Economic Strategy (SG-D-04, SG-E-01, SG-E-15, SG-E-25)

Singapore's economic strategy has always been premised on attracting foreign investment, building world-class infrastructure, and positioning the country as a hub for regional and global economic activity. AI reinforces this model: Singapore is positioning itself as the AI hub for Southeast Asia, just as it positioned itself as the financial hub, shipping hub, and petrochemical hub in earlier decades. The risk is the same risk that has always accompanied Singapore's hub strategy: dependence on external capital, technology, and talent that could relocate if conditions change.

Defence (SG-D-03, SG-F-21)

AI's military applications — autonomous systems, intelligence analysis, cyber warfare, predictive maintenance — are not explicitly addressed in the National AI Strategy, which focuses on civilian applications. But the SAF's Total Defence doctrine (see SG-D-29) and the Cyber Security Agency's expanded mandate (see SG-F-22) signal that AI's defence applications are being developed separately from the public-facing strategy.

Financial Sector (SG-D-14, SG-E-02, SG-E-18)

MAS's FEAT principles, Project Veritas, and the new AI Risk Management Guidelines make the financial sector the most actively governed domain for AI in Singapore. The DBS case study shows that the financial sector is also where AI's workforce displacement effects are appearing first and most visibly.

Labour and Manpower (SG-D-10, SG-E-19, SG-E-20, SG-E-26, SG-E-39)

The AI workforce displacement question is inseparable from Singapore's broader labour policy challenges: the foreign worker dependency, the progressive wage model, the SkillsFuture initiative, and the gig economy platform worker regulations. AI's ability to automate cognitive tasks — not just manual ones — means that the traditional strategy of moving workers "up the value chain" through training and skills upgrading may face diminishing returns if AI occupies an increasing share of the value chain.

POFMA and Information Control (SG-D-27)

AI-generated disinformation — deepfakes, synthetic text, AI-powered influence operations — represents a category of threat that POFMA was not designed to address. The Elections Amendment Act (2024) addresses one specific vector (election deepfakes), but the broader challenge of governing AI-generated content across the information ecosystem remains unresolved. The government's strengthening of public service media as a "frontline defence" against AI-driven misinformation (announced in the March 2026 Committee of Supply) signals awareness of the problem but not a comprehensive solution.

US-China Relations (SG-F-12)

Singapore's emergence as a major node in the global AI chip supply chain — and the associated risk of being drawn into US-China technology decoupling — adds a new and urgent dimension to the bilateral relationship management that has defined Singapore's foreign policy since independence. The AI chip issue is not a typical trade dispute; it involves national security, military technology, and strategic competition between Singapore's two most important economic partners.

---

12. Assessment — Singapore's AI Wager

Singapore's approach to AI can be understood as a wager with four components:

First, that AI will be the defining technology of the mid-21st century. This is the safest element of the wager. Every major government, technology company, and research institution shares this assessment. The question is not whether AI will matter, but how much.

Second, that Singapore can be an AI deployer rather than an AI creator. PM Wong stated explicitly that "our advantage does not lie in building the latest frontier models. It lies in deploying AI effectively, responsibly, and at speed." This is a realistic assessment of Singapore's position — it cannot compete with the US, China, or the EU in frontier model development — but it carries a dependency risk. If the frontier model developers (OpenAI, Google, Meta, Anthropic, Chinese labs) restrict access, increase prices, or impose conditions on their models, Singapore's deployment advantage is contingent on others' willingness to supply.

Third, that innovation-friendly governance will attract more investment and talent than prescriptive regulation would deter. This is the most contested element. Singapore's voluntary frameworks, no-fine regime, and open-door talent policy contrast sharply with the EU's legally binding AI Act. The wager is that companies will prefer to develop and deploy AI in Singapore precisely because the regulatory environment is lighter. If this wager succeeds, Singapore becomes the "AI Singapore" brand it aspires to be. If it fails — if a major AI-caused harm exposes the inadequacy of voluntary governance — the government will face pressure to retrospectively legislate, at which point the narrative of regulatory predictability collapses.

Fourth, that the workforce can be retrained fast enough to avoid sustained job displacement. The SkillsFuture ecosystem, the AI Apprenticeship Programme, the "AI Bilingual" worker concept, and the National AI Impact Programme represent a comprehensive institutional response to the displacement challenge. But the pace of AI capability advancement may outrun the pace of human retraining. DBS's Piyush Gupta — one of Singapore's most respected corporate leaders — stated flatly that for the first time in 15 years, he is struggling to create new jobs to replace those AI is eliminating. If DBS, with its S$1 billion AI programme and 1,000 new AI positions, cannot fully absorb its own displaced workers, the challenge for smaller companies with fewer resources will be more severe.

The outcome of this wager will be determined not by Singapore's strategy documents or governance frameworks, but by the speed and scope of AI capability development — a variable over which Singapore has no control. What Singapore can control is the quality of its institutional response: how quickly it adapts its education system, how effectively it retrains displaced workers, how intelligently it governs AI deployment, and how honestly it acknowledges the costs alongside the benefits. The government's track record — from industrialisation in the 1960s to digitalisation in the 2000s — suggests it will adapt. The question is whether AI's pace of change exceeds even Singapore's capacity for adaptation.

---

Appendix A: Timeline of Key AI Developments in Singapore

---

Appendix B: Key Institutions and Their AI Roles

---

Update — DC-CFA2 and the Green-Compute Mandate (December 2025)

The December 2025 launch of Singapore's Second Data Centre Call for Application (DC-CFA2) marks a material extension of the policy framework described above. Where the first DC-CFA (March 2022) re-opened the post-moratorium pipeline at approximately 60 MW under a selective approval regime, and the October 2025 Jurong Island low-carbon park designation set aside roughly 20 hectares with potential for up to 700 MW, DC-CFA2 is the first tranche under which Singapore conditions data-centre capacity allocation explicitly on the AI use case and on what is, to date, the most stringent green-energy and efficiency mandate of any major data-centre jurisdiction.

The headline parameters of DC-CFA2, as published by the Infocomm Media Development Authority and the Economic Development Board on 1 December 2025, are: approximately 200 MW of additional capacity allocated through a selective process; a Power Usage Effectiveness (PUE) ceiling of 1.25, materially below the regional peer norm of 1.4–1.6; and a requirement that successful applicants source at least 50 per cent of their power from low-carbon or renewable sources. The 1.25 PUE ceiling, in particular, places Singapore at the global frontier — only a handful of hyperscale operators in temperate climates currently achieve that figure at scale, and doing so in Singapore's tropical conditions requires liquid-cooling architectures that are still relatively novel. The 50 per cent green-energy requirement is operationally enforceable only because of the parallel build-out of cross-border clean electricity imports (the Indonesia 2 GW conditional approval and the 6 GW import target by 2035) and the hydrogen retrofit pathway documented in SG-O-06.

The strategic intent is twofold. First, DC-CFA2 attempts to resolve the structural tension that the 2019 moratorium was designed to manage: that AI demand for compute capacity is growing faster than Singapore's electricity grid can decarbonise, and that uncontrolled data-centre growth would either crowd out other electricity demand or force the grid back onto fossil-fuel generation. By making green-energy sourcing a condition of approval rather than an aspiration, the regime forces the data-centre operator to internalise the energy-transition cost rather than pass it to the grid. Second, DC-CFA2 operationalises the "sovereign AI" framing that has been articulated rhetorically since the National AI Strategy 2.0 (December 2023): if Singapore is to host frontier-AI compute on shore for governments and firms in the region — the model being executed by the Singtel-NVIDIA sovereign AI factory partnership — the compute infrastructure must satisfy environmental constraints that are themselves part of the value proposition to environmentally constrained customers.

The vulnerability that the regime does not resolve is the one identified in §3 above: Singapore's role as a chip and compute hub is itself a node in the US-China technology competition, and the strict environmental gating of new capacity does not change the underlying geopolitical exposure to export controls, end-use restrictions, or the diversion-prevention architecture that the United States has been progressively tightening since 2024. DC-CFA2's capacity allocation is, in this sense, a green-energy story sitting on top of an export-controls story — the green-energy story is the publicly emphasised one because it is the one Singapore can fully control.

Primary sources for the DC-CFA2 update (added 2026-04-30):

• Infocomm Media Development Authority and Economic Development Board, DC-CFA2 launch announcement, 1 December 2025 (coverage and analysis: Morgan Lewis, "Singapore Announces Data Center Capacity Allocation Call," March 2026, https://www.morganlewis.com/pubs/2026/03/singapore-announces-data-center-capacity-allocation-call)
• Data Center Dynamics, "Singapore opens call to develop 200MW of data center capacity," December 2025, https://www.datacenterdynamics.com/en/news/singapore-opens-call-to-develop-200mw-of-data-center-capacity/
• NVIDIA, "Singtel sovereign AI," 2025, https://blogs.nvidia.com/blog/singtel-sovereign-ai/

---

This document is part of the Singapore Governance Corpus, Block O — Emerging Themes. It should be read alongside SG-D-17 (Smart Nation), SG-D-04 (Economic Strategy), SG-E-25 (Digital Economy), and SG-F-12 (US-China Rivalry) for full context on Singapore's AI positioning.