Data centers powered by artificial intelligence are on track to consume more electricity than entire nations within two years. Global data center electricity consumption reached approximately 415 terawatt-hours in 2024, representing about 1.5 percent of the world's total electricity use. By one estimate, this figure could approach 1,050 terawatt-hours by 2026, which would make data centers the fifth-largest energy consumer globally, ranking between Japan and Russia . The growth rate tells an even more dramatic story: data center electricity consumption has been climbing at a compound annual growth rate of 12 percent since 2017, more than four times faster than overall global electricity consumption.

What's Driving This Explosive Growth in Data Center Power Demand?

The surge in energy consumption stems directly from the rapid expansion of artificial intelligence infrastructure. About 60 percent of the electricity used in data centers powers the servers, which contain central processing units (CPUs) and graphics processing units (GPUs) that process data. However, for larger hyperscaler data centers optimized specifically for AI workloads, this figure jumps to around 75 percent. These AI-focused servers use chips that consume 2 to 4 times more watts than conventional counterparts .

The International Energy Agency (IEA), in its base case scenario, projects that global data center electricity consumption could reach 945 terawatt-hours by 2030, climbing further to 1,200 terawatt-hours by 2035. Deloitte forecasts a similar trajectory, predicting a rise to 1,065 terawatt-hours by 2030. Goldman Sachs Research forecasts a 160 to 165 percent increase in power demand by 2030 compared to 2023 levels .

The impact is especially acute in the United States, which currently accounts for 45 percent of global data center electricity consumption. The IEA estimates that data center energy demand in the U.S. will increase by 130 percent by 2030 .

How Are AI Training and Inference Consuming Different Amounts of Power?

Understanding where the energy actually goes requires breaking down AI workloads into two distinct processes. Training involves running thousands of high-performance chips continuously for weeks or months to teach a model how to recognize patterns and generate responses. This process is extraordinarily energy-intensive, consuming gigawatt-hours of electricity for a single frontier model. Inference, by contrast, is the process of using a trained model to generate responses to user queries .

The numbers reveal a surprising imbalance: estimates suggest that 80 to 90 percent of computing power used for AI comes from inference, not training. In 2024, a single query on an advanced generative AI model like ChatGPT required an estimated 2.9 watt-hours of electricity, nearly 10 times the 0.3 watt-hours needed for a conventional Google search. Newer measurements suggest the median energy per text query has fallen to 0.24 to 0.3 watt-hours, although this can be much higher for long reasoning tasks or multimodal prompts that combine text, images, and other data types .

The IEA's modeling indicates that electricity consumption from servers used for AI workloads, predominantly inference, is projected to grow by 30 percent annually as adoption increases. This single category of usage is expected to account for almost half of the net increase in global data center consumption between 2024 and 2030 .

Steps to Understanding the Global Energy Policy Response

Policymakers and international governance bodies are beginning to grapple with the energy implications of AI expansion. Several key approaches are emerging to address the challenge:

• Measurement and Standards Development: International governance efforts are working to build out standardized measurement tools and reporting frameworks so that energy consumption from AI can be accurately tracked and compared across different data centers and regions.
• Energy Policy and Investment Strategies: Governments and regulatory bodies are exploring energy policies and investment approaches designed to manage energy constraints while supporting hyperscaler expansion and sustainability goals simultaneously.
• Grid Capacity Planning: Experts are examining how power availability and grid capacity may constrain future AI compute growth, requiring coordination between energy providers and technology companies on infrastructure development.

The dialogue on these issues has become urgent because the uncertainties surrounding corporate energy disclosures and inconsistencies in measurement methodologies make it difficult to assess the true scale of the problem. Nevertheless, independent projections of actual and future usage indicate that energy demand from data centers, driven primarily by AI, is projected to add hundreds of terawatt-hours to global energy consumption by 2030 .

What Does This Mean for the Future of AI Infrastructure?

The energy challenge extends beyond simple power consumption figures. The rapid growth in AI-driven data center demand is creating a mismatch between electricity generation capacity and consumption needs. This gap could become a limiting factor for AI development itself, potentially constraining how quickly companies can build new data centers and deploy new AI models at scale.

The cloud computing market for AI is expected to drive 10 to 15 percent of total cloud spending by 2030, with the infrastructure layer alone forecasted to account for 29 percent of the AI cloud market. This infrastructure layer encompasses the computing power, networking, frameworks, and services that organizations use to build foundational AI models . As this market grows, so too will the pressure on electrical grids worldwide.

The challenge is not merely technical but also geopolitical and economic. Countries and regions with abundant, affordable electricity will have competitive advantages in hosting AI data centers. Meanwhile, regions facing electricity constraints may struggle to participate in the AI economy or may face rising energy costs that ripple through their broader economies.

For now, the trajectory is clear: data centers powered by AI are becoming a dominant force in global energy consumption, and the infrastructure, policy, and investment decisions made in the next few years will shape whether this growth can be managed sustainably or whether it will strain electrical grids and climate goals worldwide.