AI Power Use in Data Centers Nearing 50%

Artificial intelligence (AI) systems are on track to consume nearly 50% of all data center power by the end of this year, according to a new analysis. This surge in energy demand highlights growing concerns about the environmental impact of AI and the need for sustainable practices in the tech industry.

The International Energy Agency (IEA) has projected that AI could require as much energy by the end of the decade as Japan uses today. This prediction underscores the urgent need for energy-efficient AI technologies and sustainable data center operations.

Energy Consumption of AI

The IEA estimates that data centers, excluding those used for cryptocurrency mining, consumed 415 terawatt hours (TWh) of electricity last year. Recent research suggests that AI already accounts for 20% of this consumption, a figure that is expected to rise significantly in the coming years.

Alex de Vries-Gao, founder of the Digiconomist tech sustainability website, based his estimates on the power consumed by chips from Nvidia, Advanced Micro Devices, and Broadcom, which are used to train and operate AI models. His findings, slated for publication in the sustainable energy journal Joule, highlight the substantial energy demands of AI systems.

Factors Influencing AI Energy Demand

Several factors contribute to the rising energy consumption of AI systems. These include the energy efficiency of data centers, the electricity consumption related to cooling systems for servers handling AI workloads, and the increasing complexity of AI models.

De Vries-Gao's research suggests that by the end of 2025, AI systems could consume nearly 49% of total data center power, excluding crypto mining. This consumption could reach 23 gigawatts (GW), which is twice the total energy consumption of the Netherlands.

Sustainability Concerns

The high energy demands of data centers, which are essential to AI technology, raise significant sustainability concerns. As AI systems become more integrated into various industries, their energy consumption is expected to grow, further straining global energy resources.

De Vries-Gao notes that several factors could influence hardware demand for AI systems, including fluctuations in demand for AI applications and geopolitical tensions that could constrain AI hardware production. For example, barriers to Chinese access to chips have led to innovations like the DeepSeek R1 AI model, which uses fewer chips and reduces computational and energy costs.

However, efficiency gains could also encourage more widespread use of AI, potentially offsetting some of the energy savings. Additionally, multiple countries developing their own AI systems could increase overall hardware demand.

Environmental Impact

The environmental impact of AI energy consumption is a growing concern. Companies like Microsoft and Google have acknowledged that their AI initiatives are challenging their ability to meet internal environmental targets. De Vries-Gao describes the AI industry as opaque, with limited information available on AI's power demands.

The EU AI Act requires AI companies to disclose the energy consumption behind training a model, but not for day-to-day use. Prof Adam Sobey of the Alan Turing Institute emphasizes the need for greater transparency regarding AI energy consumption and the potential for AI to improve the efficiency of carbon-emitting industries.

Future Outlook

As AI continues to advance, the need for sustainable energy practices in the tech industry becomes increasingly urgent. Innovations in AI hardware and energy-efficient data center operations will be crucial in mitigating the environmental impact of AI systems.

De Vries-Gao's research highlights the importance of addressing the energy demands of AI to ensure a sustainable future for the technology. With increasing awareness and action, the tech industry can work towards reducing the environmental footprint of AI while continuing to innovate and develop groundbreaking technologies.