Preloader
Others
  • Estimated reading time: 2 Minutes

Why Data Centres Are Running Short of Electricity: Serhiy Tokarev on a Global Challenge for the AI Industry

Why Data Centres Are Running Short of Electricity: Serhiy Tokarev on a Global Challenge for the AI Industry

The world's largest technology companies, which develop the most expensive graphics processors, have encountered an unexpected obstacle. It is not a dispute over AI model testing or a shortage of semiconductors. The problem is a lack of available electricity, equipment and power grid capacity. Serhiy Tokarev, investor and founder of Tokarev Foundation, explained how the shortage of energy capacity is changing the economics of building and scaling artificial intelligence infrastructure.

The Battle for Megawatts: What Experts Are Saying

According to estimates from the International Energy Agency, global digital infrastructure will require 945 TWh of electricity in 2030, compared with just 415 TWh consumed in 2024. Experts calculate that demand will increase by 15% annually, four times faster than in all other sectors of the economy combined.

Power grids are struggling to keep pace with this growth. For example, in the US, waiting times for new large transformers now exceed two years. According to Morgan Stanley, American data centres could face a power shortfall of around 49 GW by 2028. Meanwhile, Microsoft CEO Satya Nadella has pointed out that the main constraint is no longer chips but electricity.

However, the problem concerns not only the amount of electricity available but also power density. While a single server rack required 7–12 kW in previous years, NVIDIA GB200 systems now require 120–132 kW. Future generations will be even more powerful and could consume up to 600 kW.

"While higher power density requires more advanced cooling, electricity shortages are prompting technology giants to secure their own power supplies. Microsoft is financing the restart of the Three Mile Island reactor, while Google has begun collaborating with Kairos Power on small modular reactors," says the founder of Tokarev Foundation.

How Ukraine's Experience Could Help

A similar problem is emerging in the EU. The European Commission plans to build up to seven AI Gigafactories and attract around €30 billion in private and public funding. This will help triple computing capacity by 2032.

"One practical option is to draw on Ukraine's experience with distributed generation. Years of war and repeated attacks on the energy system have inevitably created a need to deploy local electricity sources in areas where the centralised grid cannot guarantee a stable supply," says Serhiy Tokarev.

The Ukrainian government has reported the installation of approximately 600 gas-fired generation units, some of which are located near water supply facilities and hospitals. They can operate independently and provide a combined capacity of more than 570 MW.

Ukraine's experience cannot simply be transferred to the global development of AI infrastructure, but it provides an example of how to respond when a centralised grid cannot supply sufficient power. The country has already begun simplifying connection and permitting procedures, while also encouraging the development of its own generation capacity.

Serhiy Tokarev emphasises that the development of artificial intelligence increasingly depends on securing sufficient megawatts of electricity. Today, simply purchasing chips is no longer enough to build operational computing capacity. Access to electricity must also be secured.

Our Sponsors

Our blog is proudly supported by industry-leading sponsors.