Singapore is easy to underestimate on a map. At the end of 2025, the entire country covered 744.3 square kilometres, a little over 47 percent of Greater London’s area. Yet in Counterpoint Research’s 2025 comparison of the world’s major AI hubs, that small island city-state finished first.

The complete top five was Singapore, Seoul, Beijing, Dubai and San Francisco. That last name makes the result feel especially counterintuitive. San Francisco sits at the centre of the modern generative-AI boom, surrounded by frontier-model companies, venture capital and research talent. So how did Singapore beat it?

The short answer is that Counterpoint’s Global AI Cities Index was not trying to identify the city that builds the most powerful model. It was asking which metropolitan ecosystems were best at turning AI into infrastructure, companies, public programmes and working applications.

On that broader test, Singapore’s size may be less a handicap than part of the explanation.

Counterpoint studied AI adoption across 100 of the world’s largest metropolitan areas. Its researchers analysed more than 5,000 initiatives in the public and private sectors, then considered communications infrastructure, data-centre and supercomputing projects, university output, startup strength and the technology vendors active in each city.

It also examined deployment across industries including healthcare, transport, telecommunications and public administration. In other words, the index was built to reward an ecosystem that can move AI out of a demonstration and into daily use.

That is different from ranking cities by academic citations, venture funding, patents, computing power or the number of frontier-model companies. San Francisco might dominate several of those narrower tables and still finish behind a place with more coordinated adoption across government and industry.

The distinction matters because “top AI city” sounds more absolute than the research supports. Singapore led this particular adoption-and-ecosystem index. It did not win every conceivable measure of artificial intelligence.

The geographic comparison in the headline is not a metaphor. Singapore’s official statistics service puts its total land area at 744.3 square kilometres at the end of December 2025. The Greater London Authority gives London’s administrative region an area of 1,572 square kilometres.

Half of Greater London would be 786 square kilometres, so Singapore fits below that mark by roughly 42 square kilometres. Land reclamation means Singapore’s measured area changes over time, but the comparison remains sound.

Smallness alone does not create an AI hub. Plenty of compact places lack the capital, talent or infrastructure to compete globally. What Singapore’s scale can do is shorten institutional distance. The national government is also, in practical terms, the government of the city. Regulators, universities, infrastructure agencies and corporate decision-makers operate within one tightly connected system.

A programme that might require national, regional and municipal agreement elsewhere can cross fewer administrative boundaries in Singapore. A pilot can be tested in a large share of the domestic market without spanning a continent. Lessons can travel quickly between public agencies and businesses because many of the relevant people are close enough to meet, coordinate and try again.

That compactness does not remove bureaucracy or disagreement. It simply makes coordination a more tractable engineering problem.

The other striking number comes from Singapore’s Economic Development Board. It says the country has attracted 80 of the world’s top 100 technology firms to establish a presence, with many using Singapore for regional or global headquarters.

“Presence” needs to be read carefully. It does not mean 80 of those companies were founded in Singapore, moved their principal headquarters there or operate a large AI research laboratory in the country. Some presences are much deeper than others. The EDB is also an investment-promotion agency, so it has an obvious reason to present the ecosystem in its strongest light.

Even with those qualifications, the concentration is consequential. AI projects rarely emerge from a single organisation working alone. A hospital may need secure cloud infrastructure and an integration partner. A manufacturer needs engineers who understand both models and factory processes. A startup needs capital, customers and access to computing. A government agency needs vendors that can work within procurement, privacy and security rules.

Put a large share of those organisations into one commercial hub and the number of possible partnerships rises quickly. The advantage is not just having famous company logos in office towers. It is having suppliers, customers, researchers and regulators within the same professional network.

Singapore launched its first national AI strategy in 2019 and National AI Strategy 2.0 in December 2023. The second plan aimed to more than triple the country’s pool of AI practitioners to 15,000, expand access to computing, deepen research and encourage adoption throughout industry and government.

The important part is not the existence of a strategy document. Almost every government now has one. Singapore paired the ambition with institutions that repeatedly turn policy into projects.

AI Singapore is a good example. Its 100 Experiments programme brings companies with a real problem and usable data together with AI engineers and apprentices. The teams spend months building a proof of concept or production-oriented minimum viable product, with co-funding reducing the risk for participating organisations. AI Singapore says the underlying method has been refined through more than 300 completed projects.

That model tries to solve two shortages at once. Companies gain help moving from an attractive AI idea to working software, while apprentices gain experience on problems that organisations genuinely need solved. The talent programme feeds the deployment programme, and the deployment programme creates more experienced talent.

This helps explain why Counterpoint highlighted Singapore’s public-private activity. The country did not depend on one spectacular laboratory. It built repeatable pathways connecting government support, universities, engineers and companies across healthcare, finance, logistics, manufacturing and public services.

San Francisco remains one of the world’s most important centres for creating AI technology. Its concentration of model developers, chip customers, founders, researchers and investors is extraordinary. Singapore finishing above it should not be read as evidence that this centre of gravity has suddenly crossed the Pacific.

The ranking instead reveals what the index chose to value. A city received credit not only for inventing AI but for deploying it across institutions, supporting it with networks and computing infrastructure, training people to use it and building partnerships that make projects repeatable.

Singapore is unusually well designed for that kind of score. It combines a capable state, a business-friendly regional hub, major universities, strong digital infrastructure and a domestic market small enough for coordinated pilots. Its government can also act as an early customer and convenor rather than waiting for a fragmented market to organise itself.

The presence of Dubai in fourth place reinforces the point. Dubai also finished ahead of San Francisco, largely because Counterpoint counted extensive government-led AI activity across its economy. This was a table of ecosystem mobilisation, not simply a census of elite research labs.

That does not make the result artificial. Getting useful systems adopted safely is a genuine capability, and one that technology rankings often neglect. It does mean readers should resist turning first place into a universal claim.

The island’s compactness helps people and institutions connect. It is far less helpful when AI needs land, electricity, cooling systems and room for high-voltage infrastructure.

Singapore already operates more than 70 cloud, enterprise and colocation data centres with over 1.4 gigawatts of capacity. Its Infocomm Media Development Authority says the Green Data Centre Roadmap is intended to provide at least another 300 megawatts in the near term, with further growth tied to cleaner energy.

Those numbers expose the harder side of becoming an AI hub. Singapore has almost no spare land, limited domestic renewable-energy options and a tropical climate that makes cooling more demanding. Every additional cluster of high-performance chips competes for physical resources in a country that must also power homes, transport and industry.

As ScienceBlog has previously examined, the environmental load of AI is not confined to the water or electricity used inside a server building. It also extends into the power system and hardware supply chain. A city can rank highly for digital readiness while exporting part of its physical footprint elsewhere.

For Singapore, resource-efficient models and hardware are therefore not pleasant additions to the strategy. They are structural requirements. The country’s long-term position will depend partly on how much useful computation it can extract from each square metre and unit of electricity.

Counterpoint Research is a commercial market-research firm. Its public overview names the broad ingredients of the index, but the complete report requires a subscription and the public material does not provide a full item-level dataset and weighting table. Outsiders can understand the logic of the ranking more easily than they can reproduce the score.

That makes this one index, not a settled verdict on the world’s AI capital. Change the weights toward frontier-model research or venture funding and San Francisco might rise. Emphasise manufacturing and hardware, and other Asian cities could gain. Measure residents’ access, privacy protection or realised productivity rather than initiatives, and the order might change again.

There is also a basic gap between activity and benefit. A city can announce laboratories, training schemes and data centres without proving that AI has made public services fairer, raised ordinary workers’ incomes or improved daily life. The strongest future version of an AI-city index would connect inputs such as investment and infrastructure to outcomes people can actually feel.

Still, Singapore’s first-place finish tells us something real. The country has compressed policy, capital, research, training, corporate operations and digital infrastructure into a remarkably small area. Its advantage is not that geography stopped mattering. It is that decades of institution-building taught Singapore how to make proximity productive.

Less than half the size of Greater London, it has assembled an AI ecosystem large enough to be compared with the world’s most famous technology capitals. The next test is whether that dense machinery produces benefits as impressive as the ranking.