Bloomberg New Economy Forum Discussion

Minister Josephine Teo’s Key Points

On Cost Barriers for SMEs

Mrs. Teo directly addressed concerns that AI adoption is prohibitively expensive for smaller businesses. Her main argument centers on the misconception that AI always requires costly infrastructure. She used a practical example of a neighborhood restaurant that deployed an AI assistant app to optimize promotions—demonstrating that not every SME needs direct access to expensive GPUs. The solution was simply an accessible app, not specialized hardware.

Singapore’s Multi-Pronged Support Strategy

Mrs. Teo outlined several approaches Singapore uses to help SMEs manage AI adoption costs:

1. Platform-based tools: Making AI capabilities available through accessible applications and platforms
2. Sector-specific initiatives: Targeted support for particular industries through AI Centres of Excellence. She cited a predictive maintenance tool developed specifically for precision manufacturers as an example of this sectoral approach.

On Global AI Capabilities

Regarding international development, Mrs. Teo emphasized that capability-building should be the primary investment area for countries. She identified three key capability dimensions:

• Skills diversity: Entrepreneurial and engineering competencies
• Stakeholder groups: Workforce, enterprises, and academia all need development
• Infrastructure investment: Countries should reshape their foundations for an AI-ready future

She acknowledged that international organizations like the World Bank, IMF, and UN Development Programme are stepping up to assist nations. Notably, she took a pragmatic stance on global AI equity, suggesting that while countries may not always match others’ achievements, “that shouldn’t hold us back from trying.”

Google DeepMind COO Lila Ibrahim’s Perspective

Accessibility Through Integration

Ms. Ibrahim’s approach focuses on embedding AI tools into existing programs to expand access rather than creating entirely new systems. This integration strategy works across Google’s various platforms to lower barriers to entry.

Multilingual and Cultural Adaptation

She highlighted Google DeepMind’s new AI research lab in Singapore, which has a specific mission: improving AI models’ understanding of Southeast Asian languages and cultural nuances. This represents a commitment to making AI genuinely accessible beyond English-speaking markets.

Global-First Design Philosophy

Ibrahim articulated a borderless approach to AI development: “We launch a model, and it’s available worldwide.” This global accessibility mindset influences their design decisions from the ground up, including:

• Multimodal capabilities from inception (as implemented in Gemini)
• Ease of interaction as a core design principle
• Literacy considerations across diverse user populations

Comparing Perspectives on Bridging the AI Divide

Areas of Alignment

Both speakers agreed on several fundamental principles:

1. Accessibility as a Core Priority

• Teo: Focus on affordable, practical applications that don’t require expensive infrastructure
• Ibrahim: Design models to be easy to interact with and available worldwide from launch

2. Localization Matters

• Teo: Sector-specific solutions tailored to local industry needs
• Ibrahim: Regional research labs focusing on local languages and cultural context

3. Multi-Stakeholder Responsibility

• Teo: Governments, international organizations, and various domestic sectors must all contribute
• Ibrahim: Tech companies should integrate AI into existing programs rather than creating parallel systems

Divergent Emphasis

Government vs. Private Sector Roles:

• Mrs. Teo emphasized governmental and institutional infrastructure—policy frameworks, sectoral support programs, and international cooperation through bodies like the World Bank
• Ms. Ibrahim focused on corporate product decisions—model design, multimodal capabilities, and technical accessibility features

Scope of Solutions:

• Teo’s approach is more structural and systemic: building national capabilities, reshaping infrastructure, and creating ecosystem-wide support mechanisms
• Ibrahim’s approach is more product-centric: ensuring the technology itself is inherently accessible through thoughtful design

Expectations for Equity:

• Teo took a realistic stance, acknowledging that not all countries will achieve the same level of AI capabilities but advocating for continued effort regardless
• Ibrahim presented a more optimistic vision of borderless technology where a single model launch instantly provides worldwide access

The Broader Context: Over 100 Countries Falling Behind

The UN report mentioned by moderator Haslinda Amin—indicating that over 100 countries, mostly from the Global South, are falling behind in AI—provides important context. This creates two complementary responses:

The Infrastructure Response (Teo): Countries need support building foundational capabilities, with help from international financial and development institutions.

The Design Response (Ibrahim): AI companies should build accessibility into their products from the start, reducing the infrastructure burden required for adoption.

Implications

The discussion reveals that bridging the AI divide requires both top-down institutional support and bottom-up accessible design. Singapore’s approach appears to blend both: government-facilitated sectoral programs combined with partnerships with companies like Google DeepMind that prioritize accessible, culturally-aware AI tools.

The most interesting tension is between Teo’s pragmatic acknowledgment that gaps may persist and Ibrahim’s vision of borderless AI access. This suggests the real challenge isn’t just technical accessibility, but the capability to effectively deploy and benefit from accessible AI tools—which returns to Teo’s emphasis on skills development across the workforce, enterprises, and academia.

Executive Summary

This case study examines Singapore’s approach to democratizing artificial intelligence adoption among small and medium enterprises (SMEs), based on insights from the November 2025 Bloomberg New Economy Forum. It explores how Singapore is addressing cost barriers, building sectoral capabilities, and positioning itself within the global AI divide affecting over 100 countries.

---

Case Study: Singapore’s Multi-Tiered AI Adoption Framework

Background Context

The Challenge:

• SMEs perceive AI adoption as prohibitively expensive, requiring costly GPUs and specialized infrastructure
• Over 100 countries, predominantly from the Global South, are falling behind in AI capabilities (UN Report, 2025)
• A new digital divide is emerging between AI-enabled and AI-excluded economies

Singapore’s Position: As a small, resource-constrained nation, Singapore cannot compete on infrastructure scale with tech superpowers. Instead, it has adopted a capability-first, access-optimized strategy that emphasizes practical deployment over raw computing power.

Case Example 1: The Neighborhood Restaurant

Scenario: A local neighborhood eatery in Singapore sought to optimize its promotional campaigns to attract customers and reduce food waste.

Traditional Perception:

• SME owner believed AI required expensive GPU infrastructure
• Estimated costs seemed prohibitive for a small F&B business
• Technical expertise appeared to be a barrier

Actual Implementation:

• Deployed an accessible AI assistant application
• No direct GPU investment required
• App-based solution integrated into existing operations
• Used AI to analyze customer patterns and optimize promotion timing

Outcomes:

• Demonstrated that AI adoption doesn’t always require capital-intensive infrastructure
• Showed practical, immediate business value for micro-enterprises
• Validated the platform-based distribution model for AI capabilities

Key Insight: The case illustrates Singapore’s principle that AI should be delivered as a service, not as infrastructure that every business must build independently.

Case Example 2: Precision Manufacturing Sector

Scenario: Singapore’s precision manufacturing industry needed predictive maintenance capabilities to reduce downtime and improve operational efficiency.

Approach:

• Sectoral AI Centre of Excellence developed industry-specific tool
• Predictive maintenance solution tailored to precision manufacturing needs
• Shared infrastructure model across multiple manufacturers

Implementation Model: Rather than requiring each manufacturer to develop its own AI solution, Singapore created a collective capability through sector-focused innovation hubs.

Strategic Rationale:

• Pools resources for SMEs with similar needs
• Builds domain-specific expertise
• Creates economies of scale in AI development
• Reduces individual company risk and investment

Outcomes:

• Industry-wide capability uplift
• Shared learning and best practices
• Lower per-company adoption costs
• Faster deployment timeline

Singapore’s Three-Pillar Support Framework

Pillar 1: Platform-Based Tools

Approach: Make AI accessible through applications and platforms rather than requiring infrastructure ownership

Mechanisms:

• Government-facilitated partnerships with tech companies
• App marketplaces for SME-focused AI solutions
• Cloud-based AI services with pay-per-use models

Target Beneficiaries: Micro and small enterprises with limited technical capacity

Pillar 2: Sectoral AI Centres of Excellence

Approach: Develop industry-specific AI solutions through collaborative innovation hubs

Mechanisms:

• Industry-government-academia partnerships
• Shared R&D for common sectoral challenges
• Knowledge transfer programs
• Pilot project funding

Target Beneficiaries: Industry clusters and medium-sized enterprises

Pillar 3: Capability Development

Approach: Build human capacity across the entire AI ecosystem

Mechanisms:

• Workforce training programs (entrepreneurial and engineering skills)
• Enterprise capability-building initiatives
• Academic research partnerships
• International collaboration for knowledge exchange

Target Beneficiaries: All stakeholder groups—individuals, companies, and institutions

International Dimension: Google DeepMind Singapore Lab

Strategic Partnership: Google DeepMind’s establishment of an AI research lab in Singapore represents a symbiotic relationship:

For Google DeepMind:

• Access to Southeast Asian linguistic and cultural expertise
• Regional hub for testing multilingual AI models
• Collaboration with Singapore’s research ecosystem

For Singapore:

• Technology transfer and knowledge spillovers
• Training ground for local AI talent
• Enhanced attractiveness for other tech investments
• Regional leadership positioning

Research Focus:

• Southeast Asian language understanding
• Cultural nuance recognition
• Multimodal AI accessibility
• Regional context adaptation

Broader Impact: The lab exemplifies Singapore’s strategy of attracting global players while building domestic capabilities—a model that smaller nations can potentially replicate.

---

Outlook: Future Trajectories and Implications

Short-Term Outlook (2025-2027)

For Singapore

Accelerated SME Adoption

• Expect 40-60% of Singapore SMEs to deploy at least one AI application by 2027
• Platform-based tools will dominate early adoption
• F&B, retail, and logistics sectors likely to lead

Sectoral Deepening

• Expansion of AI Centres of Excellence to additional industries
• Manufacturing, healthcare, and professional services as next focus areas
• Development of Singapore-specific AI solutions for tropical climate challenges (construction, agriculture)

Talent Pipeline Maturation

• First cohorts from AI-focused education programs entering workforce
• Growing pool of “AI-literate” middle managers in SMEs
• Emergence of local AI consultancy sector serving regional SMEs

Regional Hub Consolidation

• Singapore positioned as Southeast Asia’s AI intermediary
• Translation layer between global AI models and regional needs
• Increasing AI service exports to neighboring countries

For Global AI Divide

Widening Before Narrowing

• Gap between AI-capable and AI-excluded nations will initially expand
• First-mover advantages in AI will compound through 2026-2027
• Countries without coherent AI strategies will face competitive disadvantage

International Institution Response

• World Bank, IMF, and UNDP scaling up AI capacity-building programs
• Emergence of “AI for Development” funding mechanisms
• Focus on foundational digital infrastructure in developing nations

Platform Accessibility Improvements

• Major tech companies expanding multilingual capabilities
• Reduced computational requirements for inference (smaller models)
• Edge AI enabling offline capabilities in low-connectivity regions

Medium-Term Outlook (2027-2030)

Technology Evolution

Democratization of AI Tools

• Continued reduction in AI deployment costs
• “No-code” AI platforms enabling non-technical users
• Industry-specific AI app stores emerging
• Edge computing reducing cloud dependency

Capability Distribution

• Open-source models challenging proprietary systems
• Regional AI champions emerging in major markets (India, Brazil, Indonesia)
• Specialized models for low-resource languages gaining traction

Economic Restructuring

SME Transformation

• AI becoming baseline expectation, not competitive advantage
• Consolidation among SMEs unable to adapt
• New categories of “AI-native” micro-enterprises
• Changing skill requirements across all business sizes

Labor Market Shifts

• Displacement of routine cognitive and analytical tasks
• Growth in “AI training and supervision” roles
• Premium on uniquely human skills (creativity, empathy, complex judgment)
• Workforce reskilling becoming continuous necessity

Productivity Divergence

• 30-50% productivity gap between AI-adopting and non-adopting SMEs
• Industry concentration as AI-enabled companies scale faster
• Geographic clustering around AI capability hubs

Geopolitical Developments

Regional Blocs

• Southeast Asian AI cooperation framework likely by 2028
• Shared infrastructure and capability-building initiatives
• Data governance harmonization attempts
• Tension between regional cooperation and great power competition

Sovereignty Concerns

• Countries seeking “AI independence” through domestic model development
• Data localization requirements affecting AI service delivery
• National security considerations constraining technology transfer
• Balancing openness with strategic autonomy

Long-Term Outlook (2030-2035)

Systemic Transformation

The “AI-Embedded Economy”

• AI infrastructure becomes invisible, like electricity
• Every business interaction mediated by AI assistance
• Real-time optimization of supply chains, pricing, and resource allocation
• Shift from “adopting AI” to “operating in AI environment”

New Divides Emerge

• Beyond the current “have/have-not” split
• Quality divide: sophistication of AI implementation
• Agency divide: who controls the AI systems
• Benefit divide: who captures economic value created by AI

Singapore’s Evolved Position

• From AI adoption hub to AI governance model
• Regulatory frameworks exported to other jurisdictions
• Mediator between competing AI governance paradigms (US, China, EU)
• Test bed for human-AI collaboration models

Critical Uncertainties

1. Will “Accessible AI” Remain Accessible?

• Risk: Frontier models require increasing computational resources
• Counter-trend: Efficiency improvements in model architecture
• Key factor: Open-source vs. proprietary model dynamics

2. Can Capability-Building Keep Pace?

• Challenge: AI advancing faster than workforce adaptation
• Risk: Structural unemployment in routine cognitive work
• Mitigation: Continuous learning systems and social safety nets

• \

3. Will Sectoral Models Suffice?

• Question: Can industry-specific solutions scale globally?
• Alternative: Winner-take-all dynamics favor platform monopolies
• Outcome depends on: Regulatory intervention and interoperability standards

4. How Will Data Governance Evolve?

• Tension: Data sharing needed for AI training vs. privacy/sovereignty concerns
• Impact: Could fragment global AI ecosystem into incompatible blocs
• Singapore’s role: Potential bridge between governance models

---

Strategic Implications

For Policymakers

1. Infrastructure Investment Must Be Strategic Singapore’s model shows that capability-building trumps raw infrastructure spending for most nations. Rather than attempting to build GPU farms to compete with tech giants, focus on:

• Developing human capital
• Creating sectoral innovation hubs
• Facilitating access to global AI platforms
• Building regulatory and governance frameworks

2. SME Support Requires Tiered Approaches One-size-fits-all policies fail because:

• Micro-enterprises need turnkey solutions (apps)
• Small businesses benefit from sectoral tools
• Medium enterprises can invest in customization
• Policy must address all three simultaneously

3. International Cooperation Is Essential No country can solve the AI divide alone. Singapore’s success depends partly on:

• Partnerships with tech companies (Google DeepMind)
• Regional collaboration (Southeast Asian focus)
• Engagement with international institutions
• Technology transfer arrangements

4. Timing Matters The 2025-2027 window represents a critical period. Countries that establish AI foundations now will be better positioned for the 2030s. Delays compound disadvantages.

For Business Leaders

1. The AI Adoption Decision Is Binary Businesses that delay AI adoption risk:

• 30-50% productivity disadvantage by 2030
• Loss of competitive positioning
• Inability to attract talent
• Customer migration to AI-enabled competitors

2. Start Small, Scale Systematically Singapore’s restaurant example validates the approach:

• Begin with specific, high-value use cases
• Use platform-based tools to minimize upfront investment
• Build organizational AI literacy gradually
• Scale complexity as capabilities grow

3. Sector Collaboration Reduces Risk Participating in industry-wide initiatives:

• Shares development costs
• Accelerates learning
• Establishes industry standards
• Reduces individual company risk

4. Human Capital Is the Constraint Technology is increasingly accessible; capability is not. Priority investments:

• Training existing workforce in AI tools
• Hiring for AI literacy, not just technical skills
• Building organizational change capacity
• Creating culture of experimentation

For Technology Companies

1. Accessibility Must Be Designed-In Google DeepMind’s approach shows that waiting to “adapt” technology for developing markets fails. Instead:

• Multilingual capabilities from inception
• Multimodal interfaces for low-literacy users
• Offline and low-bandwidth operation modes
• Cultural context awareness

2. Platform Models Beat Custom Development For SME markets, success comes from:

• Turnkey solutions requiring minimal configuration
• Pay-per-use pricing models
• Industry-specific templates and workflows
• Support ecosystems (consultants, trainers, integrators)

3. Regional Hubs Create Network Effects Singapore lab model suggests value of:

• Local presence for credibility and customization
• Regional expertise centers
• Collaboration with local institutions
• Knowledge transfer to domestic ecosystems

For International Development Organizations

1. AI Must Become Core Development Priority The World Bank, IMF, and UNDP escalation signals recognition that AI capability gaps will overwhelm traditional development interventions. Countries lacking AI capabilities will face:

• Declining export competitiveness
• Inability to participate in global value chains
• Brain drain to AI-capable economies
• Widening income gaps

2. Infrastructure Approach Needs Updating Traditional development focus on physical infrastructure must expand to:

• Digital connectivity as foundational layer
• Cloud computing access
• Data center partnerships
• AI platform subscriptions as development aid

3. Capability-Building Requires Long-Term Commitment Unlike traditional infrastructure projects, AI capability development:

• Takes 5-10 years to mature
• Requires continuous investment, not one-time projects
• Depends on ecosystem development, not individual interventions
• Needs private sector partnership from inception

---

Risks and Challenges

Implementation Risks

1. SME Resistance and Inertia

• Many SMEs remain skeptical of AI value proposition
• Change management capacity limited in small firms
• Fear of workforce displacement creates resistance
• Short-term focus prevents long-term investment

Mitigation: Demonstrator projects, subsidized pilots, peer learning networks

2. Capability Gap Persistence

• Training programs may not produce needed skills fast enough
• Brain drain of AI talent to higher-paying markets
• Generational divide in AI literacy
• Mismatch between education outputs and industry needs

Mitigation: Continuous learning systems, immigration policy, industry-education partnerships

3. Dependency on Foreign Platforms

• Over-reliance on Google, Microsoft, Amazon, etc.
• Geopolitical tensions could disrupt access
• Data sovereignty concerns
• Limited domestic AI development capacity

Mitigation: Multi-vendor strategies, open-source alternatives, strategic autonomy investments

Systemic Risks

1. Winner-Take-All Dynamics

• AI advantages compound rapidly
• Early adopters capture disproportionate benefits
• Late movers face insurmountable gaps
• Regional and global inequality increases

Potential Outcome: Despite Singapore’s success, most developing nations fall further behind, creating a two-tier global economy with limited mobility between tiers.

2. Labor Market Disruption

• AI displaces routine cognitive work faster than new jobs emerge
• SME workers face higher displacement risk than large company employees
• Reskilling at scale proves infeasible
• Social instability from technological unemployment

Potential Outcome: Political backlash against AI adoption, protectionist policies, slowdown in technology deployment.

3. Concentration of Economic Power

• AI platform providers become gatekeepers
• SMEs lack bargaining power in platform ecosystems
• Data extraction benefits platforms, not data generators
• National economies lose policy autonomy

Potential Outcome: Regulatory intervention, antitrust action, push for open-source alternatives, geopolitical fragmentation of AI ecosystem.

4. Governance Failure

• AI development outpaces regulatory capacity
• Competing governance models (US, EU, China) fragment ecosystem
• Lack of international coordination
• Ethical concerns unaddressed (bias, privacy, accountability)

Potential Outcome: Crisis event triggers reactive, restrictive policies that stifle beneficial innovation along with harmful applications.

---

Recommendations

For Countries Seeking to Replicate Singapore’s Model

1. Assess Starting Position Realistically

• Singapore has advantages (wealth, education, governance capacity) most countries lack
• Adapt the principles, not the specific mechanisms
• Focus on relative improvement, not absolute parity

2. Prioritize Sectors Strategically

• Identify industries with highest economic impact
• Choose sectors where AI tools already exist
• Start with areas having willing early adopters
• Build momentum through visible successes

3. Leverage Regional Partnerships

• No single small country can go it alone
• Create regional sectoral hubs (different countries focus on different industries)
• Share infrastructure and capability-building costs
• Establish common standards and data governance

4. Balance Openness and Strategic Autonomy

• Welcome foreign AI companies but with capability transfer requirements
• Use open-source models to reduce dependency
• Develop domestic talent even while importing expertise
• Maintain optionality in platform choices

For Global AI Governance

1. Establish “AI Access” as Development Right

• Create international framework recognizing AI access as essential for economic participation
• Develop metrics for AI capability gaps
• Set targets for reducing global AI divide
• Create accountability mechanisms

2. Fund AI Infrastructure as Global Public Good

• International development banks finance AI platform access
• Subsidized cloud computing for developing nations
• Open-source model development funding
• Multilingual AI development grants

3. Create Technology Transfer Mechanisms

• Incentivize AI companies to establish regional hubs
• Require capability-building in exchange for market access
• Facilitate knowledge sharing between countries
• Protect intellectual property while enabling learning

4. Build Governance Interoperability

• Harmonize AI regulations where possible
• Create mutual recognition frameworks
• Establish data flow agreements
• Prevent fragmentation into incompatible blocs

---

Conclusion

Singapore’s approach to AI adoption for SMEs offers a pragmatic model that balances ambition with realism. By focusing on accessibility over infrastructure ownership, capability-building over technology acquisition, and sectoral collaboration over individual company efforts, Singapore demonstrates that smaller nations can participate meaningfully in the AI economy without attempting to match superpowers in raw computational resources.

However, the case also reveals the limits of national action. Singapore’s success depends partly on factors beyond policy control: partnerships with companies like Google DeepMind, integration with global AI platforms, and participation in international development efforts. The UN finding that over 100 countries are falling behind suggests that without coordinated international action, the AI divide will widen despite best national efforts.

The outlook through 2035 presents a bifurcated future. In the optimistic scenario, AI accessibility continues improving, capability-building efforts mature, and international cooperation narrows the divide. In the pessimistic scenario, AI advantages compound, winner-take-all dynamics dominate, and most countries find themselves permanently excluded from the AI economy’s benefits.

Singapore’s model won’t solve the global AI divide alone, but it offers principles that can scale:

• Practical over perfect: Start with accessible applications, not ambitious infrastructure
• Collective over individual: Build sectoral capabilities, not just company-level solutions
• Continuous over one-time: Treat capability-building as ongoing process, not project
• Collaborative over competitive: Engage with global platforms while building domestic capacity

The next 3-5 years will determine whether these principles can be deployed widely enough to prevent the permanent bifurcation of the global economy into AI-enabled and AI-excluded zones. Minister Josephine Teo’s closing remark—”that shouldn’t hold us back from trying”—captures the essential posture: pragmatic determination in the face of daunting challenges.

The question is not whether AI will transform the global economy. It will. The question is whether that transformation will be broadly shared or narrowly concentrated—and the answer depends on choices being made right now.

Maxthon

In an age where the digital world is in constant flux and our interactions online are ever-evolving, the importance of prioritising individuals as they navigate the expansive internet cannot be overstated. The myriad of elements that shape our online experiences calls for a thoughtful approach to selecting web browsers—one that places a premium on security and user privacy. Amidst the multitude of browsers vying for users’ loyalty, Maxthon emerges as a standout choice, providing a trustworthy solution to these pressing concerns, all without any cost to the user.

Maxthon, with its advanced features, boasts a comprehensive suite of built-in tools designed to enhance your online privacy. Among these tools are a highly effective ad blocker and a range of anti-tracking mechanisms, each meticulously crafted to fortify your digital sanctuary. This browser has carved out a niche for itself, particularly with its seamless compatibility with Windows 11, further solidifying its reputation in an increasingly competitive market.

In a crowded landscape of web browsers, Maxthon has forged a distinct identity through its unwavering dedication to offering a secure and private browsing experience. Fully aware of the myriad threats lurking in the vast expanse of cyberspace, Maxthon works tirelessly to safeguard your personal information. Utilizing state-of-the-art encryption technology, it ensures that your sensitive data remains protected and confidential throughout your online adventures.

What truly sets Maxthon apart is its commitment to enhancing user privacy during every moment spent online. Each feature of this browser has been meticulously designed with the user’s privacy in mind. Its powerful ad-blocking capabilities work diligently to eliminate unwanted advertisements, while its comprehensive anti-tracking measures effectively reduce the presence of invasive scripts that could disrupt your browsing enjoyment. As a result, users can traverse the web with newfound confidence and safety.

Moreover, Maxthon’s incognito mode provides an extra layer of security, granting users enhanced anonymity while engaging in their online pursuits. This specialised mode not only conceals your browsing habits but also ensures that your digital footprint remains minimal, allowing for an unobtrusive and liberating internet experience. With Maxthon as your ally in the digital realm, you can explore the vastness of the internet with peace of mind, knowing that your privacy is being prioritised every step of the way.