Kyrgyzstan emerges as a global leader in low-cost AI computing

Yesterday 15:00

By: Azzat Manal

Kyrgyzstan emerges as a global leader in low-cost AI computing

Kyrgyzstan could offer some of the lowest-cost artificial intelligence computing in the world, according to a World Bank analysis that compares electricity costs without taking government subsidies into account.

The finding highlights the potential economic advantages of countries with abundant and relatively inexpensive energy resources as demand for AI infrastructure continues to expand. The World Bank noted that Kyrgyzstan ranks first in the study when electricity is valued at its underlying cost rather than subsidized consumer prices.

The analysis, based on research by World Bank economist Mikhail Lokshin, examined 85 countries and estimated the cost of operating an Nvidia H100 SXM AI accelerator. In Kyrgyzstan, the estimated hourly cost was $1.58, placing the Central Asian country at the top of the ranking.

Ethiopia followed closely at $1.59 per hour, while Kosovo, Canada, Tajikistan and Montenegro were estimated at around $1.60. The results suggest that electricity prices can play a decisive role in determining the competitiveness of AI computing facilities, particularly in energy-intensive data centers.

Countries combining hydropower resources with cooler climates appear to have a particular advantage. Lower electricity costs can reduce one of the largest operating expenses for data centers, while colder weather can also help limit the energy required for cooling high-performance computing equipment.

However, the World Bank stressed that the figures represent theoretical potential rather than an existing AI infrastructure advantage. Several countries in the Europe-Central Asia region appear highly competitive on paper, but have yet to develop significant data-center capacity.

Kyrgyzstan, for example, currently has only around 5 megawatts of installed data-center capacity. Expanding that infrastructure would require substantial investment in computing equipment, electricity networks, telecommunications and reliable data-center facilities.

The potential economic returns could nevertheless be significant. A preliminary estimate cited by the report suggests that a 40-megawatt data center in Kyrgyzstan could generate between $630 million and $950 million in annual gross revenue under favorable market conditions.

Such figures would be particularly significant for a relatively small economy, potentially making AI infrastructure an important source of export revenue and investment. The country would still need to overcome challenges related to financing, infrastructure, connectivity and the availability of advanced computing equipment before turning its theoretical energy advantage into a major AI industry.

The findings underline a broader shift in the global AI race, in which access to affordable and reliable electricity is becoming almost as important as access to computing technology itself. As demand for AI services grows, countries with low-cost energy and suitable infrastructure could increasingly compete for data-center investment and a larger share of the expanding digital economy.