AI Is Eating Data Center Power Demand, Only Getting Worse

AI Is Eating Data Center Power Demand, Only Getting Worse

Artificial intelligence (AI) is rapidly becoming one of the largest consumers of global data center energy. Recent analysis warning that its power demands could double by year’s end. According to new research published in Joule by Alex de Vries-Gao, AI already accounts for up to 20% of global data center electricity use. Additionally, it could rise to nearly 50% by the end of 2025—excluding bitcoin mining. 

This surge is driving by the explosive adoption of generative AI models like ChatGPT, which require immense computational power. Compared to the relatively modest investments in cryptocurrency mining, the capital poured into AI by tech giants such as Google and Microsoft represents a far more aggressive and energy-intensive trajectory. 

Google, for instance, has seen its greenhouse gas emissions rise 48% since 2019, largely due to increased AI compute demands, complicating its net-zero commitments. The International Energy Agency (IEA) estimates global data centers used 415 terawatt-hours (TWh) of electricity in 2024—about the same as Saudi Arabia’s total annual energy use. This figure is projected to exceed 900 TWh by 2030, with AI being a key contributor. 

De Vries-Gao’s methodology analyzes the AI supply chain—focusing on semiconductor manufacturing capacity at Taiwan Semiconductor Manufacturing Company (TSMC), which produces high-performance chips for Nvidia, Google, and others. His estimate suggests AI could consume up to 82 TWh in 2025, equivalent to Switzerland’s annual electricity use. 

However, researchers warn of data gaps. Without clearer transparency from tech firms on AI energy metrics, estimates remain broad. Experts like Sasha Luccioni of Hugging Face stress the need for disclosure to ensure sustainable AI development. 

As AI reshapes global IT infrastructure, managing its energy impact becomes a pressing priority for enterprise sustainability strategies. 

 

Source: 

https://www.wired.com/story/new-research-energy-electricity-artificial-intelligence-ai/  

Get Started

Ready to Build Your Next Product?

Start with a 30-min discovery call. We'll map your technical landscape and recommend an engineering approach.

000 +

Engineers

Full-stack, AI/ML, and domain specialists

00 %

Client Retention

Multi-year partnerships with global enterprises

0 -wk

Avg Ramp

Full team deployed and productive