At the southern tip of Malaysia lies the state of Johor, renowned for its beaches and mountainous jungle. But Johor has a recent industry boom: generative AI-enabled data centers Microsoft allocating over $2 billion for just such a data center. For tech giants, electricity has become the recent oil. A state-of-the-art AI data center might need 90 MW, enough to power tens of thousands of American homes. As AI applications proliferate, from chatbots to AI agents, the needs grow. One industry consortium plans data centers requiring 10 GW of power (over a hundred times more than today’s largest). Providing low-cost and reliable energy is now as critical to technology companies as silicon chips.
In 2025, substantial tech companies will be scouring the world for kilowatts, megawatts and gigawatts. During board meetings, discussions about server performance are increasingly overshadowed by discussions about network capacity and the future of energy. Nations blessed with an abundance of low-cost energy are using this newfound advantage and craft policies to attract AI investment with a fervor once reserved for manufacturing.
Regions that have made a name for themselves in the past in data centers, such as Ireland and Singapore, found their capacity stretched to the limit before the GenAI boom. This has created opportunities for unexpected competitors, not only Malaysia but also Indonesia, Thailand, Vietnam and Chile. The delay is less critical than maintaining the flow of electrons.
Affordable energy has long been a priority for companies: just as companies in the past located their refineries near ports and factories near coal mines, artificial intelligence companies are trying to locate near places where they can constantly obtain electricity – and this at favorable prices. prices.
Location ultimately matters. Half of the energy costs in a data center typically come from running cooling and air conditioning systems that keep servers from overheating. Colder climates or coastal areas will become more sought after as potential locations.
This pull to deliver AI is so powerful that substantial tech companies are buying filthy energy to meet it, putting their own and local economies at risk. decarbonization goals at risk.
Countries compete fiercely for data center operations. Tax breaks are popular: More than half of US states – including Arizona, Up-to-date York and Texas – offer operators some form of tax breaks and even preferential rates for purchasing land and committing to energy access. in Malaysia, Green Belt Pathway initiatives speed construction permits, cutting red tape and enabling accelerated construction – and power lines – for data centers. Concessions on data rules allowing the free flow of information.
This interplay of watts and algorithms is redrawing the map of global influence. This is a change as profound as the oil boom of the 20th century, but much less noticeable. No pipelines are being built, no tankers are changing course. Instead, inconspicuous warehouses filled with servers are becoming recent geopolitical flashpoints.
It is unclear to what extent this changes global influence. Real AI research – where breakthroughs happen – will remain in research centers in San Francisco, London, Beijing and Paris. However, the data centers that bring these algorithms to market will be low-margin, high-cost, low-selling businesses.
Electrodiplomacy will be a key pillar for the next few years. Scaling AI is less about algorithms and more about electronics.
However, nations seizing this moment should exercise caution; their advantage may prove short-lived as dominant economies figure out how to bring low-cost, neat energy online in sufficient quantities to encourage domestic hosting.
For today’s power-hungry AI data center providers, the challenge is to turn this fleeting advantage into a sustainable advantage. They will need to move beyond attracting data centers and build their own sustainable innovation ecosystems that can thrive long after the energy rush has ended.