Sergey Tokarev on the Energy Challenge Facing AI: Why Technology Companies Are Competing for Megawatts
The world’s largest technology companies are investing billions of dollars in graphics processors as they race to build artificial intelligence infrastructure, but they are increasingly facing difficulties connecting them to power supplies. Data centres lack sufficient grid capacity, electricity and essential equipment. The International Energy Agency projects that global electricity consumption by data centres will grow by around 15% annually between 2024 and 2030, reaching approximately 945 TWh in 2030, compared with 415 TWh in 2024.
Sergey Tokarev, investor, IT entrepreneur and founder of Tokarev Foundation, explained why access to energy resources is becoming just as important as the availability of chips.
Why Power Grids Cannot Keep Pace with AI Development
Energy infrastructure requires significant upgrades and expansion because it cannot keep up with the needs of technology companies. In the US, delivery times for large transformers previously ranged from 12 to 16 weeks, but now exceed two years. Morgan Stanley forecasts that the power shortfall for American data centres could reach 49 GW by 2028.
One way to supply data centres with electricity more quickly is to build gas-fired power plants nearby. However, turbine manufacturers are also overwhelmed with orders. GE Vernova has already secured contracts for equipment totalling 116 GW, while its annual production capacity stands at 20 GW.
“Technology giants are increasingly focusing on developing their own power generation capacity. For example, Google has reached an agreement with Kairos Power on the use of small modular reactors, while Microsoft has begun financing the restart of the Three Mile Island nuclear reactor to purchase its energy,” says Sergey Tokarev.
How Ukraine’s Experience with Distributed Generation Could Help the World
Distributed generation is one possible solution to electricity shortages. As Russian forces continue to attack Ukraine’s energy infrastructure, the country has rapidly developed this sector. As a result, approximately 600 small gas-fired generation units with a combined capacity of 570 MW have already been installed. Some of these facilities keep hospitals and water utilities operating when the central power grid experiences disruptions.
Ukraine’s distributed generation capacity has already reached 1.8 GW, with plans to install a further 1.5 GW in the future.
According to Sergey Tokarev, Ukraine’s experience with distributed generation cannot simply be transferred to the global AI industry. However, it demonstrates how energy storage systems and local power plants can improve the resilience of overall infrastructure.
How Access to Electricity Affects Investment in Artificial Intelligence
European Union countries are facing a similar challenge. They plan to triple the EU’s computing capacity by 2032. To achieve this, up to seven AI Gigafactories will be established, supported by more than €30 billion in public and private funding. However, electricity is needed now.
Energy infrastructure must be planned as an integrated system. It should include:
- Installing transformers and connecting to the power grid.
- Providing sufficient cooling for high energy consumption.
- Developing on-site or local power generation capacity.
- Installing energy storage and backup power systems.
“The global AI race is becoming increasingly dependent on a stable electricity supply. Today, the winners will be those who can provide enough power for data centres equipped with modern technology,” emphasises founder of Tokarev Foundation.