From 18 to 21 June 2025, the CAMCA Regional Forum convened in Ulaanbaatar, Mongolia. CAMCA \u2014 Central Asia, Mongolia, Caucasus, Afghanistan \u2014 is a non-political discussion platform spanning ten countries: Afghanistan, Armenia, Azerbaijan, Georgia, Kazakhstan, Kyrgyzstan, Mongolia, Tajikistan, Turkmenistan, and Uzbekistan. Aziz Soltobaev moderated the plenary on Geopolitical Implications of AI Models in the CAMCA Region<\/em>.<\/p>\n\n\n\n The framing question was where the ten countries actually sit in the AI infrastructure map. Per Epoch AI’s reporting, 75 per cent of AI supercomputers are located in the United States and 15 per cent in China. The historically familiar compute powers \u2014 the United Kingdom, Germany, Japan \u2014 now play a notably smaller role. At the same time, around 80 per cent of total compute capacity sits inside the largest private technology companies, up from roughly 40 per cent seven years earlier. The location of a supercomputer does not determine who uses it \u2014 cloud access redistributes capacity \u2014 but cross-border access for governments, universities, and companies in other countries is shaped by the geopolitical relationship between their state and the state where the compute physically sits.<\/p>\n\n\n\n The panel discussion turned on three changes already underway in the AI-producing countries: shifts in the labour market, shifts in where private capital is concentrating, and the social changes that follow when both move at once. The question for the CAMCA ten was practical: how to build the geopolitical relationships that keep AI infrastructure access open, and what to weigh first when drafting national AI policies that have to function under those constraints.<\/p>\n\n\n\n