The AI Energy Dilemma: When the Digital Boom Overloads the Grids

Artificial intelligence has become an ever-hungry power guzzler. Experts warn that the global electricity consumption of AI data centers is expected to increase elevenfold between 2023 and 2030 – from 50 to around 550 billion kWh. If non-specialized data centers are included, the total consumption for central data processing in 2030 is expected to be 1400 billion kWh. For comparison: a single modern data center with 100 MW can account for the electricity needs of around 100,000 households annually. Future facilities could require up to 20 times more energy.

The high and rapidly growing energy demand of AI infrastructure is increasingly pushing local power grids to their limits, writes the civil society organization AlgorithmWatch in a new analysis. Cities like Frankfurt/M., which have become hotspots for the data center industry, are facing a strained supply situation.

In fact, data center operators have already secured all available power capacities in the Main region, says Max Kendl from the Frankfurt Chamber of Industry and Commerce (IHK). Although new registrations in the grid area of the Main metropolis are generally still possible, the NGO explains, the implementation of further capacities requires long-term lead times. The energy provider Mainova is preparing for the surge with additional grid connection points to the transmission network, new power lines of all voltage levels, and additional or modernized substations.

According to the research, this scarcity means that around 20 percent of planned data center projects in Europe are threatened by delays or complete halts. The necessary power connection is simply lacking. The problem is global: even tech giants in the USA cannot utilize billions worth of AI chips because the power supply is insufficient.

Reliance on Natural Gas

The expansion of data centers has profound ecological and economic consequences, it is stated. Despite the shift towards renewable energies, data centers will continue to rely on fossil fuels such as natural gas in the coming years. This comes with high ecological costs. The use of fossil fuels is becoming an answer for data centers – as recently also in Frankfurt/M. – to the energy bottlenecks they themselves cause. The Federal Environment Agency warns that rapid growth is at the expense of the climate. Some operators are relocating to countries with less stringent regulations to circumvent climate requirements.

The increase in electricity consumption by data centers is associated with a rise in greenhouse gas emissions. In addition, there is a massively increasing demand for water for cooling – an almost fourfold increase to 664 billion liters is expected by 2030 – as well as millions of tons of additional electronic waste.

According to the report, rising electricity bills are also linked to the explosive demand from data centers for AI, cloud computing, and streaming services. In the USA, electricity prices have risen by 34 percent from 2020 to August 2025. Analysts fear that electricity costs could explode worldwide due to AI's energy hunger. In Germany, it is predicted that the annual electricity consumption of data centers will more than double from 20 billion kWh by 2030, which could bring new dangers of increased end-customer prices.

Moving to the Outskirts

For reasons of operational reliability, cloud service providers are often forced to build three or more data centers in a region. As space in hotspots like Frankfurt/M. becomes scarce, the "outskirts" for data centers are expanding into surrounding regions. A large planned project by NTT in Nierstein near Frankfurt/M., for example, is set to have a capacity of 480 MW from 2026, which corresponds to the electricity demand of about 500,000 households. Google recently announced the construction of a "state-of-the-art cloud data center" in Dietzenbach, Hesse, and the expansion of the existing one in Hanau for several billion euros. Critics estimate that in the banking capital itself, the 75 hectares designated as suitable areas are overvalued until 2030.

To overcome the challenges, strategic and long-term planning of the grid infrastructure with a horizon of 20 to 25 years is essential. Researchers are calling for intelligent grids that can flexibly respond to peak loads, integrate storage, and be conceived across sectors. Concepts such as AI Energy Hubs, which link data centers with local power generation and energy management, are being discussed as part of the solution. At the same time, environmental organizations warn that AI's electricity hunger can only be managed through a consistent expansion of renewables.

(akn)