Electricity demand for AI data centers will reach 2.5 GW by 2030

With the growing popularity of neural networks, there is an increasing need for data center (DC) capacity, which is driving up colocation prices due to higher electricity tariffs and problems connecting to power grids. By 2042, according to forecasts, the share of mining, DCs, and artificial intelligence (AI) could account for 27% of the country's total electricity consumption.

According to the International Energy Agency, by 2030 electricity demand from DCs will more than double, with AI development as the main driver. By 2027, Russia's big data and AI market will grow to 644 billion rubles. Electricity consumption by data processing centers will reach 2.5 GW, experts believe.

DCs optimized for AI are characterized by high energy consumption. Special racks designed for graphics processing units (GPUs) require from 50 to 150 kW, and in the future their consumption may reach 300 kW.

Against the backdrop of rising electricity prices and capacity shortages, the cost of colocation in DCs has increased since 2022 by 11% in the Moscow region and by 19% in St. Petersburg. This is forcing companies to look for ways to optimize their costs.

According to market experts, the problem is not only the cost of electricity but also the difficulty of connection. Studies show that more than 75% of DCs are concentrated in Moscow, where a capacity shortage is predicted. To cover demand by 2030, more than 7 GW of new generation capacity will need to be built.

In light of these challenges, many companies, such as Yandex, continue to invest in expanding their data centers, seeking to adapt to the conditions imposed by AI development. However, most experts consider mass construction of AI-focused DCs impractical until market requirements stabilize.