Central Asia is emerging as an unexpected front in the global buildout of AI computing power, according to a report by Yevgeniya Mikhailidi for Nikkei Asia, datelined Tashkent and Istanbul.
Nikkei Asia describes the race to build data centers in the region as intensifying, with two projects setting the early pace.
Uzbekistan is moving first but small. Its TAS-1 facility, rated at 6 megawatts, is due to be completed by the end of this year — a modest footprint by global standards, but a signal that the country intends to host AI infrastructure rather than rent it from abroad.
Kazakhstan is aiming considerably higher. Nikkei reports a planned 125-megawatt facility stocked with 100,000 Nvidia chips, targeted for 2027. That is roughly twenty times the power capacity of Uzbekistan's first site, and the chip count puts it in the conversation with serious commercial AI clusters.
The megawatt figures matter more than they might appear. Data center capacity is measured in electricity because power, not floor space, is the binding constraint on how many AI chips a site can actually run. A jump from 6MW to 125MW is a jump in national ambition — and in the grid investment required to support it.
The contrast between the two projects also captures the shape of the competition: Uzbekistan getting something operational quickly, Kazakhstan betting on scale over speed. Neither country has historically been a hub for the kind of computing infrastructure that AI training demands.
Why it matters: where AI data centers get built determines which countries control the computing power behind modern AI — and which merely buy access to it — so a race in Central Asia is a bid by two nations to own a piece of that infrastructure rather than depend permanently on someone else's.