The global expansion of the Artificial Intelligence (AI) market is generating a sharply rising energy demand. Analyses estimate the market volume in 2023 at 190 billion US dollars, with a projection to nearly five trillion US dollars by 2033; global spending on AI could already exceed 2.5 trillion dollars by 2026. This development goes hand in hand with a massive increase in the energy consumption of digital infrastructure, particularly data centres.
Markus Voigt, Executive Chairman of the aream Group, points out that renewable energies play a crucial role in the proliferation of this future technology. UN analyses demonstrate that data centres already consumed approximately 450 terawatt-hours of electricity in 2025. This consumption could rise to 1,000 TWh by 2030 and up to 1,300 TWh in subsequent years, meaning the electricity consumption of data centres would surpass entire national economies.
Training modern AI models proves to be particularly energy-intensive. GPT-4, according to the UN, is estimated to have required over 100 days and up to 70 GWh of electricity, which corresponds to 50 times the consumption of GPT-3. Future systems are likely to significantly exceed these values, thereby causing substantial CO2 emissions unless energy sources are adapted. Against this background, Mr Voigt emphasises the necessity of an accelerated expansion of renewable energies as a central prerequisite for sustainable AI growth.
Wind and solar energy currently represent the fastest-growing power sources for data centres. They are expected to be able to cover nearly 50 per cent of the additional energy demand by 2030. This development is significantly promoted by long-term power purchase agreements (PPAs) between technology companies and energy producers. To secure competitiveness in the AI era and simultaneously achieve climate targets, consistent investments in renewable energies, grid infrastructure, and storage technologies are required, according to Voigt.
The performance of assets in the aream portfolio in June illustrates the challenges of energy generation. German solar plants were affected by curtailments, a prolonged grid shutdown, and inverter malfunctions, resulting in a target achievement of 79 per cent. Including compensated volumes, the target achievement was 97 per cent. Spanish plants reached 67 per cent, primarily due to curtailments, with 1,850 MWh being reimbursed. Italian plants, however, showed stable performance with 99 per cent target achievement. Wind availability met expectations, but a better result was prevented by curtailments and plant malfunctions, leading to a target achievement of approximately 88 per cent.
