The scramble to build the physical backbone of artificial intelligence is playing out on several continents at once, and the footprint is starting to draw attention.
In South Korea, telecom operator SKT has committed to building hyperscale AI data centers in the Yeongnam region, according to Let's Data Science. Hyperscale facilities are the largest class of data center, designed to house the massive computing clusters that train and run AI systems.
The hardware buildout is reaching smaller markets too. Data Center Dynamics reports that Eleveight AI has deployed 512 Nvidia B300 chips at a 2-megawatt data center in Gagarin, Armenia — a sign that the specialized processors powering AI are spreading well beyond the usual tech hubs.
But the expansion comes with growing costs and pushback. In Texas, proposed power plants tied to AI data centers could emit 287 million tons of greenhouse gases each year, according to a report cited by KERA News. That figure underscores how much electricity the AI boom demands, and how much of it may come from fossil fuels.
The political response is building in parallel. Tri-State Alert reports that local regulations on data centers continue to crop up, even as OpenAI faces increasing federal scrutiny — a sign that governments at multiple levels are beginning to weigh in on where and how these facilities get built.
Why it matters: The infrastructure behind AI is being built faster than the rules and power grids meant to support it, setting up a collision between rapid technological ambition and real-world limits on energy and public tolerance.