From America to Asia, AI data centres are beginning to influence elections

It is becoming increasingly clear that the data centre is becoming the nuclear power station of the 21st century: everyone knows it’s needed, but no one wants one in their own back garden. The difference is that it took decades for nuclear power to become a political taboo. For data centres, it took just two years.

In the United States, the shift in sentiment amongst Republicans has come with surprising speed. In 2024, Senator Jon Husted of Ohio hailed the arrival of a Google project as proof that his state would become a technology hub. This year, with the mid-term elections just around the corner, he has tabled a bill to protect citizens from the costs of new facilities. The Washington Post has traced this trend by analysing thousands of posts and newsletters from politicians of both parties: over the past six months, the Republican tone on data centres has become markedly more critical, mirroring a shift that had already taken place amongst Democrats about a year earlier. The Republican candidate for governor of Wisconsin, Tom Tiffany, has promised to abolish tax breaks for data centres. The Republican challenger in Pennsylvania, Stacy Garrity, accuses the incumbent governor, Josh Shapiro – who is standing for a second term – of having been the sector’s biggest supporter for years, only to then announce new restrictions. A poll by the Annenberg Public Policy Centre indicates that almost half of Americans are strongly opposed to any new construction in their area. This is significant, as it means that AI and data centres have begun to sway votes.

It would be convenient to dismiss the whole affair as American electoral folklore. But the same pattern is repeating itself, albeit with different nuances, in Asia. Malaysia has long since put a freeze on new applications for data centres lacking an artificial intelligence component, whilst courting the very same investors it would like to filter out. Alibaba and Tencent have poured capital into Johor Bahru, the city overlooking Singapore; they have been joined by the Chinese group behind DayOne, the former international arm of GDS Holdings, with an investment of 3.5 billion dollars. According to Reuters, the Malaysian government has already attracted around $35 billion to the sector. Tencent Cloud has just opened its first Malaysian region, presenting it as a potential hub for the company’s entire south-eastern expansion. Indonesia and Thailand are following a similar trajectory, with Beijing and Washington vying for the same sites equipped with sufficient power supply. Kuala Lumpur is seeking the capital and geopolitical recognition that come with AI infrastructure, whilst fearing that the population’s electricity and water supplies will end up being consumed by servers housing models trained elsewhere.

In truth, the industry has largely brought much of the local opposition upon itself. Companies realised too late that a refurbished leisure centre is not enough to win over a community that sees a warehouse the size of a third of the town springing up within a matter of months. Local resistance can be partly defused through targeted investment. Regional or national resistance is another matter entirely: there, the data centre becomes a tangible symbol of foreign money and decisions taken over the heads of voters.

Hence the rush towards ‘friendly’ sites. Australia, with its land and energy reserves, has already become a testing ground for rapid IT development. This trend, known as ‘friendshoring’, is an elegant solution from a technical perspective: it dilutes the domestic political cost and distributes the geopolitical cost across a network of allies. It only holds up, however, as long as those allies trust those who promise privileged access to the most advanced models in exchange for hospitality. A last-minute export control, imposed without warning, would quickly teach foreign partners to be wary of the promises they have received.

In short, every government wants artificial intelligence infrastructure on its own territory, as proof of its relevance in a technological race that nobody wants to watch from the sidelines; yet few citizens are willing to have that infrastructure installed right on their doorstep, where the electricity bill is likely to rise long before the first promised highly specialised job arrives. Much of the politics surrounding artificial intelligence is already playing out between these two conflicting desires. The novelty is that today, fate is decided less in research laboratories and more in local councils in Malaysia and in some remote towns in Ohio.