Artificial intelligence increasingly looks like a contest designed for very large countries. US private investment in AI reached almost $286 billion last year, more than 23 times the figure recorded for China, according to Stanford University's 2026 AI Index.
Training and running advanced models increasingly demands vast computing infrastructure, expensive chips, electricity and a relatively small pool of highly skilled people.
Yet two countries with populations and domestic markets far smaller than either superpower have placed themselves unusually close to the centre of that race.
The UAE is pouring capital into computing infrastructure, models and international technology partnerships. Singapore has spent years building digital government, AI governance, research capability and one of the world's most concentrated pools of AI talent.
Stay up to date with the latest news. Follow KT on WhatsApp Channels.
Neither can realistically replicate the entire American or Chinese technology ecosystem. That, according to Spring Fu, Managing Director of The Academy at Mohamed bin Zayed University of Artificial Intelligence, may actually be the wrong objective.
“A mid-sized country can't expect to lead on every front,” Fu said. “What it can do is turn its size into an advantage in a few specific areas.”
For smaller states, she argues, that means moving government faster, concentrating resources on selected capabilities and becoming valuable partners within an AI ecosystem that no country entirely controls.
You cannot win everywhere
The UAE's most visible bet is on computing power. Stargate UAE is planned as a one-gigawatt computing cluster inside a wider five-gigawatt UAE-US AI campus in Abu Dhabi. G42 is building the project, with OpenAI and Oracle set to operate it and Nvidia supplying advanced chips. Its first 200-megawatt phase is expected to come online this year.
Singapore faces a different physical reality. It already has more than 1.4 gigawatts of data-centre capacity, but the city-state has limited land and a tropical climate that makes cooling energy-intensive. Data centres account for around 7 per cent of national electricity consumption, according to its Infocomm Media Development Authority. Rather than pursuing unrestricted expansion, Singapore has been selecting new capacity partly on energy efficiency and strategic value.
The comparison shows why there may be no single formula for becoming an AI power. Abu Dhabi can use land, energy, capital and geopolitical partnerships to develop enormous computing infrastructure. Singapore has had to extract more value from tighter physical constraints.
What both rely on heavily is access to the outside world. “For a small state, international partnerships are not a supplement to national capability; they are part of how that capability is built,” Fu said, pointing to access to chips, joint research and attracting major international companies.
That dependence is not necessarily a weakness. Even the world's largest AI ecosystems depend on international supply chains for chips, manufacturing equipment, energy, models and expertise.
The question for smaller states becomes which capabilities they need to control themselves and which they can reliably obtain through partners.
Compute can be bought, people take longer
Fu believes the harder constraint is human. “Compute can be bought or built, and capital can be deployed quickly,” she said. “But without people who can use that infrastructure, run AI systems inside organisations, and make informed decisions as regulators and buyers, the investment sits idle.”
Her definition of AI talent extends beyond a handful of elite researchers. A country needs those researchers, she said, but also far larger numbers of engineers and practitioners able to deploy the technology, followed by another group that receives much less attention: leaders who understand AI well enough to decide where it should and should not be used.
Singapore offers one version of that strategy. It had the world's second-highest concentration of AI talent last year, with 1.82 per cent of LinkedIn members classified as AI talent in Stanford's index. It is now seeking to train 100,000 workers to become “AI bilingual”, combining knowledge of their existing profession with the ability to use AI, while helping 10,000 businesses deepen adoption over the next three years.
The UAE is moving quickly in the same direction. Stanford found that its concentration of AI talent grew 121 per cent between 2019 and 2025, the fastest increase among countries measured. The federal government is also working with MBZUAI to build agentic AI expertise across 80,000 government employees, including senior leadership and non-technical roles.
Fu's own programme offers a smaller example of that third layer. Twelve UAE leaders spent five days in Singapore in September studying how the country approaches AI policy, industry, governance and talent, meeting government entities, multinationals and local companies. The objective was not to train them as engineers but to make them better decision-makers about the technology.
That exchange is becoming part of a wider relationship. Days later, the Abu Dhabi-Singapore Joint Forum identified AI and capacity-building among the areas where cooperation between the two sides will expand.
Turning activity into Capability
Perhaps the harder question is how to know when all that investment has produced something durable. “Activity is easy to generate: announcements, pilots, start-ups, and conference partnerships signed,” Fu said. “Capability is harder to see and harder to fake.”
Her tests are more demanding. Is AI working inside the core operations of government and major industries rather than remaining inside innovation labs? Can a country develop and retain the people it needs? Are procurement rules, data systems, regulators and testing standards mature enough for deployment at scale? And can those capabilities survive changes in technology, leadership and funding?
Singapore has increasingly focused on that less visible institutional layer. Its National AI Strategy was refreshed again this year, while a National AI Council chaired by Prime Minister Lawrence Wong was established in February.
Its AI governance work has also moved into agentic systems, with a framework updated this year using feedback from more than 60 organisations and examples of agents already operating inside companies and government.
Abu Dhabi, meanwhile, is pushing AI deeper into government itself. Last month, its Executive Council used an agentic AI platform for the first time to support analysis and discussion during a meeting chaired by Sheikh Khaled bin Mohamed bin Zayed Al Nahyan, Crown Prince of Abu Dhabi.
That is where the smaller-state argument becomes more interesting than simply counting data centres or investment announcements.
Neither Singapore nor the UAE can match the United States in private capital or China in industrial scale. Their wager is that they do not need to.
For smaller countries, AI power may instead depend on choosing carefully what to build, developing enough people to use it, moving institutions quickly enough to deploy it and maintaining relationships that provide access to everything else.
A data centre without talent is infrastructure. Talent without compute is potential. Partnerships without the institutions to use them remain announcements. The harder achievement is turning all three into capabilities.