Last month, Trinidad and Tobago signed agreements with American tech companies paving the way for them to build data centers in the island nation — the first deal of its kind in the Caribbean. Government officials hailed it as a transformational investment that could position the nation of just 1.5 million at the forefront of the AI boom. But not everyone was convinced.
Almost immediately, critics questioned what Trinidad was getting in return. The announcement said that it would bring over 5,000 jobs, but many in the country pointed to the enormous amounts of electricity and water that modern data centers can require. Could a smaller nation like Trinidad and Tobago keep up with the hefty energy and resource demands data centers require?
Around the world, governments and corporations – mostly from the US and China – are scrambling to build data centers, the infrastructure needed to power artificial intelligence tools. For smaller countries searching for their next engine of economic growth, that demand presents an opportunity. Just as smaller nations once competed to attract banks and offshore finance through favorable tax policies, a new race may be emerging to attract something far more concrete: the physical infrastructure that powers AI. And countries in the Caribbean, Central America, and Southeast Asia are looking to cash in.
Why would you build a data center so far away? At first glance, building a data center thousands of miles away from the North American AI companies who rely on them might be counterintuitive – building in the US would appear to be easier. The reason, though, is that not all AI computing needs to happen in your backyard.
Heather West, a geotechnology expert at Eurasia Group, explains that there are two main types of data centers. One is designed to train proprietary AI models – like Anthropic’s Claude or Google’s Gemini – a process that requires enormous amounts of computing power and electricity but can happen almost anywhere. “You run your job, the computer chugs at it until you have your results or your AI model,” she said. Once the AI is trained, the model can be deployed elsewhere closer to users.
The second type of data centers handle what’s known as AI inference, the everyday task of responding to users’ prompts with an already trained AI program, like how one might use ChatGPT. Those facilities need to be much closer to the people using them because even small delays matter, West explained.
That distinction is changing where companies look to build. For AI training centers, geography is often less important than access to abundant electricity. In many cases, the cheapest and fastest place to build isn’t Silicon Valley or elsewhere in the US – where input prices like electricity and labor may be high – but wherever companies can secure reliable power, enough land, and the internet to connect those computers to the rest of the world.
Those factors help explain Trinidad and Tobago’s appeal. Beyond its natural gas reserves, which generate nearly all of the country’s electricity, it sits at a crossroads of multiple subsea fiber-optic cables, which transmit data, internet traffic, and telecommunications signals across the globe, linking North America, South America, and the Caribbean.
Trinidad and Tobago is far from the only small nation betting on AI data centers. Mauritius, an island nation off the coast of East Africa less than half the size of Trinidad and Tobago, for example, now hosts 10 smaller telecom data centers and is investing heavily in increased connectivity with more undersea cables to expand its ability to host larger AI centers.
Malaysia has emerged as a major destination for AI data centers in Asia, recycling land once used for palm oil production, and has attracted investments from Microsoft, Google, and Amazon to build their own data centers. Some are for training AI models, but others can serve the computing needs just over the border in neighboring Singapore, where land and power have become increasingly scarce and expensive.
Costa Rica, meanwhile, hopes to capitalize on its green energy credentials. Because it generates 98% of its electricity from renewable sources, it can offer companies a significantly lower carbon footprint with data centers built in the country.
The US wants to build abroad. The United States has already begun encouraging American companies to expand AI infrastructure overseas through initiatives like the American AI Exports Program, which aims to promote US AI technology hardware, software, and data systems abroad. “Washington hopes that countries building on the American AI stack will become embedded in American tech standards and commercial ecosystems,” says Jessica Brandt, senior fellow for Technology and National Security at the Council on Foreign Relations. Once countries adopt American AI infrastructure, she adds, switching to Chinese alternatives becomes politically and technically more difficult.
China is pursuing a similar strategy. Through investments across the Global South, Beijing is financing digital infrastructure that mirrors its broader Belt and Road ambitions, according to West.
Kenya is one of those countries hosting Chinese-built data centers, and is building a 5,000-acre “technopolis” just outside of Nairobi with Chinese funding. Another Chinese firm, Range IDC, committed $5 billion to build its first data center in Batam, Indonesia, the first for the company outside of mainland China. Central Asian countries too want in on the action, although limited energy resources and connectivity have limited Chinese investment so far.
This demand – and competition between the US and China – present both an opportunity and a risk. A major AI data center can bring billions of dollars in investment, new infrastructure, and closer ties with the world’s leading AI companies. But these facilities also consume enormous amounts of electricity and resources to keep them running. For smaller countries, the challenge will be ensuring they become more than just landlords of the machines behind the AI boom.
