Technology giants are acquiring increasing numbers of graphics processors for AI, but they cannot always put them into operation. The main reason is a shortage of electricity, connection equipment and grid capacity. Microsoft states that the main constraint on artificial intelligence development today is not a shortage of chips, but access to electricity. This shortage could persist until at least the end of 2026. Serhiy Tokarev, an IT entrepreneur, investor and founder of Tokarev Foundation, explained what options are available to address the situation.
How the Energy Crisis Is Affecting the Global AI Industry
The International Energy Agency forecasts that electricity consumption by global data centre infrastructure will increase from 415 TWh in 2024 to approximately 945 TWh in 2030. Demand is expected to grow by 15% annually. The main problem is that power grids cannot keep pace with the needs of artificial intelligence.
For example, in the US, large transformers previously took 12–16 weeks to manufacture, whereas today the waiting time is around two years. According to Morgan Stanley, the power shortfall for US data centres alone could reach 49 GW by 2028.
“Building gas-fired power plants near data centres could help reduce the electricity shortage more quickly, but turbine manufacturers are also overwhelmed with orders. Production schedules are already almost fully booked through 2030,” says Serhiy Tokarev.
The issue is not only a shortage of electricity but also power density. A modern NVIDIA GB200 system consumes around 120–132 kW, compared with a conventional server rack that previously required 7–12 kW. This also creates challenges for air cooling, which is no longer sufficient. As a result, technology companies such as Microsoft and Google have already begun investing in their own power generation capacity.
Distributed Generation in Ukraine: What Lessons Can the World Learn?
EU countries plan to establish up to seven AI Gigafactories and attract more than €30 billion in funding. However, these new facilities are already facing electricity shortages.
The founder of Tokarev Foundation highlights Ukraine’s experience. Following repeated attacks on its energy system, the country began developing distributed generation. Today, Ukraine has around 600 small gas-fired generation units with a combined capacity of 570 MW. Its distributed generation capacity has already reached 1.8 GW, with another 1.5 GW planned for the future.
“In Ukraine, local power sources keep critical infrastructure, such as hospitals and water utilities, operating even when the central grid goes down. Of course, this experience cannot simply be replicated across the global AI industry. However, other countries can recognise the benefits of local generation, rapid grid connections and energy storage systems,” adds Serhiy Tokarev.
In the race to develop artificial intelligence, the advantage will go to those who can turn available megawatts into operational computing capacity the fastest.








