Resource consumption of AI: The insatiable industry and its costs

Artificial intelligence (AI) now appears in a wide variety of applications and devices – sometimes integrated in purposeful, meaningful ways, but often deployed as little more than a marketing ploy, or even imposed on users by the software. The AI boom is funneling ever more resources into the AI industry. From chip manufacturing and the expansion and operation of data centers to the training and everyday use of AI models, these systems increasingly demand vast and growing quantities of electricity, water and raw materials.

This surge is driving up global energy demand. The current and projected electricity use of data centers makes the trend unmistakably clear. The International Energy Agency (IEA) estimates that data centers currently account for only 1.5 % of global electricity consumption, but this figure is expected to rise sharply. In Germany, according to government data, they already account for more than 4% of the country’s electricity consumption. In cities such as Frankfurt am Main, where data centers are densely clustered, they already draw roughly 40% of the local power supply.

The trajectory points steeply upwards: worldwide, electricity demand could double by 2030. In Germany, the Bundesnetzagentur (Federal Network Agency) projects that data centers could account for as much as 10% of national electricity consumption by 2037. This prompts an obvious question: where will all this electricity come from?

Will climate protection be undermined by AI’s enormous energy appetite?

If no action is taken, the electricity consumption of AI will lead to fossil-fuel or nuclear power stations running for longer – at the expense of climate protection and the energy transition. Data centers already run largely on fossil fuels, and a growing number of these AI factories are now even building their own gas-fired power stations. Even when data centers claim to run on “green electricity”, this often means that renewable energy is diverted away from neighboring businesses. If fossil fuel power plants are used to supply electricity elsewhere while data centers consume all available renewable energy, it does nothing to support the energy transition – quite the contrary.

The fundamental problem is that AI data centers generate additional electricity demand, making it harder for other industries to transition to renewable energies. Therefore, AlgorithmWatch demands that new data centers should only be built if new, additional renewable energy plants are constructed at the same time.

The majority of people agree with us: data centers should only be built if additional renewable energy capacity is created to accommodate them. This is the finding of a representative survey conducted in several European countries, in which AlgorithmWatch also participated. Click here for the survey results.

Combating climate killers with clear rules

There is still very little publicly available information about the energy consumption of AI systems and the emissions they cause. The AI industry goes to considerable lengths to conceal its electricity, water and resource consumption, which makes it difficult to develop effective political strategies to reduce emissions. AlgorithmWatch has long called for greater transparency and stronger regulatory mechanisms regarding the resource consumption and environmental impact of AI systems. Climate protection and environmental protection must play a much greater role in the further expansion of infrastructure for the AI industry.

sustAIn - The Sustainability Index for Artificial Intelligence

As part of the sustAIn project, AlgorithmWatch developed practical tools to assess the sustainability of AI systems in real-world settings. We have demonstrated how AI can be made more sustainable and provided corresponding policy recommendations. Based on a set of criteria and indicators we developed to measure the environmental, social and economic sustainability costs of AI, the assessment tool is available on the sustAIn website – as is the sustAIn magazine which explores the project’s findings in greater detail.

Do you believe that climate protection should play a key role in the AI industry? Then consider becoming a supporting member of AlgorithmWatch!

With your monthly donations, you can support rigorous research and effective campaigns. In this way, you will help us hold decision-makers accountable and ensure that our expertise drives meaningful change.

More information on your supporting membership here.