Schneider Electric has cautioned policymakers to carefully guide the electricity consumption by AI data centers and prevent it from spiraling out of hand.
This comes as AI data centers’ energy consumption is reportedly continuing to increase due to surging demand for AI services, creating scope for AI companies to look for alternative sources of energy.
Data centers may leave everyone else in total darkness
Schneider Electric in their report proffered four potential scenarios and suggested some guiding principles to follow that might prevent AI data centers from “eating” the power grid and leaving the world in darkness.
The study follows the IEA Global Conference on Energy and AI that was held last month. The study, titled Artificial Intelligence and Electricity: A System Dynamics Approach examines the emerging schools of thought relating to AI and its impact on energy consumption.
While a lot has been reported on generative AI and electricity consumption, the Schneider Electric report also concurs with previous studies that existing data center infrastructure needs significant electricity to function, therefore it will require more resources to support the projected surge in AI adoption.
The anticipated increased demand for AI services and the subsequent increase in energy consumption has also caused concerns over the potential strain the technology will pose on electricity grids. There are also worries over the possible environmental impact if energy demand continues to rise at this rate.
Director of Schneider Electric Sustainability Research Institute Rémi Paccou said the study is meant to explore potential futures and prepare stakeholders to navigate the challenges and opportunities ahead.
“Instead, we hope it serves as a starting point for informed discussion and decision-making.”
Paccou.
“We present our findings with the understanding that AI is a rapidly evolving field and that our knowledge is constantly growing,” he added.
Schneider Electric therefore came up with four different scenarios and these are Sustainable AI, Limits to Growth, Abundance Without Boundaries, and Energy Crisis.
Schneider Electric projects increasing electricity demand from now to 2030
According to the study, all four scenarios that Schneider Electric came up with point to an increase in energy consumption during the period 2025 to 2030 as demand continues to surge. However, they diverge markedly based on some assumptions that underpin each scenario.
With Sustainable AI, the Schneider study looks at the potential outcomes of prioritizing efficiency while consumption rises while Limits to Growth look at a constrained path where AI development hits human-related limits. Sustainable AI offers a more promising approach that would see electricity consumption increase from an expected 100 terawatt-hours (TWh) in 2025 to 785 TWh in 2035, according to its model.
Generative AI inferencing will be the key driver of electricity consumption in the AI sector under this scenario from 2027 to 2028. There will also be a move towards more efficient and less energy-intensive models.
According to the report, it is “characterized by a symbiotic relationship between AI infrastructure and demand, where efficiency and resource conservation are mutually reinforced.”
Other scenarios such as Abundance Without Boundaries look at the potential risks of unchecked growth while the Energy Crisis looks at how the uneven energy demand and generation could lead to widespread shortages.
Total AI energy, according to the report, will increase from the baseline 100 TWh this year to 510 TWh by 2030, but challenges like manufacturing jams for specialized chips and shortage of data for LLMs taking their toll.
The report further states that the Abundance Without Boundaries scenario reflects that the ongoing rapid development of AI will create challenges as AI firms race towards bigger and more advanced infrastructure, outpacing the capacity for sustainable utilization of resources.
The Energy Crisis scenario sees rapid AI growth resulting in its energy need conflicting with other critical sectors of the economy, leading to some operational challenges for AI-dependent industries.
Under this scenario, energy consumption is projected to reach its peak in 2029, reaching about 670 TWh before dropping to 380 TWh by 2032 and another drop in 2025 to 190 TWh.
Schneider offers some suggestions for the potential energy crisis
According to the report, an uncoordinated governance results in fragmented policies which could result in fragmented policies. These will result in global or localized energy deficits.
However, the Schneider report offers some recommendations for sustainable AI and these look at three areas – AI infrastructure, AI development, governance, standards, and education.
AI infrastructure pushes that next-generation data centers should be optimized with the latest cooling technologies, high-density compute, and modern energy-efficient hardware such as GPUs and TPUs.
This also follows reports that data AI data centers are consuming large volumes of water to cool off AI servers, with tech firms like Google, Microsoft, and OpenAI reportedly seeing an increase in utility consumption at their data centers.
The recommendation under AI development suggests making models more efficient through techniques like model pruning, quantization, and lightweight architecture.
Under governance, standards, and education, the report recommends that policymakers develop and implement certification schemes for sustainable AI practices like energy efficiency and environmental impact. A robust framework will also guide responsible AI development and address energy consumption, data privacy, and ethical considerations.
From Zero to Web3 Pro: Your 90-Day Career Launch Plan
“This post has helped me solve my issue, thanks a ton!”