Most countries in the global AI race sell one of two things. The United States and China sell frontier models and the compute to train them. The Gulf states sell sovereign capital and gigawatts of power. Singapore sells neither. It has no energy surplus, a resident population smaller than many Chinese cities and a land area of about 728 square kilometers, smaller than New York City. What it built instead is harder to copy: a reputation as the best-governed and most neutral place in Asia to build and run AI.

In May 2026 that positioning produced its clearest payoff yet. At the ATx Summit in Singapore, OpenAI announced OpenAI for Singapore, a partnership with the Ministry of Digital Development and Information backed by more than S$300 million. At its center sits the company’s first Applied AI Lab anywhere outside the United States. OpenAI has not opened an applied lab in London, Tokyo or Dubai. Singapore got the first one. At the same summit, Nvidia confirmed its first research lab in the country, and Google DeepMind deepened a national partnership it had begun months earlier. Three of the most consequential names in AI, all anchoring in a country with no model of its own at the frontier.

This is the story of how Singapore got there, what the ecosystem actually contains, and the constraints that could still cap its ambitions.

Singapore planned its way into AI deliberately, over more than a decade, and the paper trail is unusually clear. The groundwork was the Smart Nation initiative, launched in 2014 as a whole-of-government push to wire public services, transport and health around data. In 2019 the country published its first National AI Strategy, choosing a handful of national projects in areas like education, healthcare and border clearance to prove AI could deliver in the real world rather than in a lab.

The arrival of ChatGPT reset the ambition. In December 2023, then Deputy Prime Minister Lawrence Wong launched the second National AI Strategy, branded “AI for the Public Good, for Singapore and the World.” NAIS 2.0 moved the country from a set of flagship projects to a systems approach built around three systems, ten enablers and fifteen actions, and it reframed AI from a useful opportunity to a strategic necessity. In the 2024 budget cycle the government committed more than S$1 billion over five years to back the plan, covering compute, talent and industry adoption.

By 2026 the strategy had matured into governance machinery. In February the government set up a National AI Council chaired by Prime Minister Lawrence Wong. In May, at the ATx Summit, it released an update with ten refreshed priorities. Minister for Digital Development and Information Josephine Teo called it a double-click rather than a system reboot, with the real work now being the translation of strategy into four National AI Missions covering connectivity, advanced manufacturing, healthcare and finance.

If there is one thing Singapore exports that almost no other mid-sized country can match, it is AI governance that the rest of the world actually uses. In January 2019 the Personal Data Protection Commission released the first Model AI Governance Framework, among the first practical guidance anywhere. In May 2022 the Infocomm Media Development Authority released AI Verify, a testing toolkit built with companies including AWS, DBS, Google, Meta, Microsoft, Singapore Airlines and Standard Chartered. In 2023 it became the AI Verify Foundation, opening the work to the global open-source community.

The generative wave extended the model. The Model AI Governance Framework for Generative AI, published on 30 May 2024, set out nine dimensions and drew input from more than 70 global organizations, including OpenAI, Google, Microsoft and Anthropic. In January 2026 the IMDA added a Model AI Governance Framework for Agentic AI, a first-of-its-kind attempt to govern AI agents while keeping humans accountable.

The story has a proof point that travels. In February 2025, Changi Airport Group received an ISO/IEC 42001 certificate for its AI management system, issued by SGS and accredited by the Singapore Accreditation Council. SGS described it as the world’s first such certificate for airport customer services, covering five passenger and commercial AI applications.

Rather than competing to build the best model, Singapore made itself the place that writes the rules others adopt.

Singapore does build models, but with a clear sense of where its leverage lies. The flagship is SEA-LION, short for Southeast Asian Languages in One Network, developed by AI Singapore. AISG dates to 2017, is supported by the National Research Foundation and is hosted by the National University of Singapore and Nanyang Technological University.

The project advanced sharply in 2025. In August it shipped SEA-LION v4, its first multimodal version, built on Google’s Gemma 3 (27B) and small enough to run on a consumer laptop. In November a second v4 model, built on Alibaba’s Qwen3-32B and supported by Alibaba Cloud, topped the regional SEA-HELM benchmark among open models under 200 billion parameters. Both are open source and distributed freely, including on Hugging Face and Google Cloud. SEA-LION is developed under the National Multimodal LLM Programme, an IMDA and National Research Foundation initiative.

Two things define both the strength and the limit. The strength is reach: earlier versions already feed downstream systems across the region, including Sahabat-AI, an Indonesian initiative that GoTo helped build on top of SEA-LION. The limit is provenance. SEA-LION is an adaptation layer on open foundation models from Google and Alibaba, not a sovereign model trained from scratch. That is a defensible choice for a small country, but it is a different proposition from the Gulf’s from-scratch Arabic models, and it should not be sold as the same thing. Deployment is real: NCS was the first to put SEA-LION v4 into production, in an internal assistant used by more than 10,000 of its staff across Asia-Pacific.

The most interesting feature of Singapore’s ecosystem is who chose to build there. In less than a year, four of the most important AI organizations on earth established or deepened a presence in the country, and the same word runs through all of them: first.

The timing is strategic. It is what a country earns when it offers regulatory clarity, political neutrality, deep English-language talent and trusted institutions in a region where AI demand is growing fast and the alternatives are less stable.

Singapore has no Mistral or Anthropic of its own. What it has is a layered base of applied AI companies, plus a willingness to host the frontier players it cannot grow itself. Crunchbase counts more than 1,400 AI companies headquartered in Singapore that have together raised over US$20 billion, yet the average one was founded only around 2019 and most are still early stage. Switch tiers below.

The thinnest part of the ecosystem is its homegrown frontier layer, and the asterisk tier shows why: some winners scale and leave, and some “Singapore” AI companies are really just based there. Singapore is less focused on building the smartest model than on being the best place to put models, talent and companies to work.

Three constraints could still cap Singapore’s ambitions.

The Green Data Centre Roadmap, launched in May 2024, tries to square the circle by trading efficiency for capacity, targeting at least 300 megawatts of additional capacity and demanding power usage effectiveness below 1.3 for new builds. A pilot in 2023 awarded about 80 megawatts to Equinix, GDS, Microsoft and an AirTrunk-ByteDance consortium, and a second call for at least 200 megawatts closed in March 2026.

Pull the ecosystem apart and the same three forces show up at every layer. The state is the architect: from Smart Nation to NAIS 2.0 to the National AI Council, almost everything traces back to a public agency setting direction. Governance is the product: AI Verify, the generative and agentic frameworks and the Changi certificate are the country’s most exportable assets. And neutrality is the strategy: as the US and Chinese AI blocs harden, Singapore has made itself the rare place where every camp will still share a room.

Whether that model can survive a genuine US-China rupture, scale its homegrown company layer beyond a handful of scaled names, and keep importing the talent it needs are the three questions the next 18 to 24 months will answer.

For now the result is hard to argue with. A country with no oil, no frontier model and little spare land convinced the most important AI company in the world to build its first overseas lab on its shores.