AI servers, especially GPU-accelerated systems, are driving rapid growth in data center electricity use, with global consumption projected to nearly double by 2030.

Current Energy Use

AI workloads are highly energy-intensive due to the computational demands of training and running large models. In 2023, AI-specific GPU servers consumed over 40 TWh globally, up from less than 2 TWh in 2017, reflecting a dramatic increase in energy demand . In the United States, data centers consumed about 4.4% of national electricity in 2024, with AI workloads being a major driver . Globally, data centers accounted for roughly 1.5% of electricity consumption in 2024, or around 415 TWh, with AI accelerating the deployment of high-performance servers .

Breakdown of Energy Use

Data center electricity consumption is distributed across several components:

• Compute servers: The largest share, especially GPU-accelerated AI servers.
• Storage systems: About 5% of total electricity, used for centralized data storage and backup .
• Networking equipment: Up to 5%, including switches, routers, and load balancers .
• Cooling and environmental control: Varies from 7% in efficient hyperscale centers to over 30% in less-efficient enterprise centers .
• Backup power systems: UPS batteries and generators, rarely used but essential for reliability .

Per-Query and Model Energy

For inference tasks, a typical ChatGPT-style query consumes roughly 0.3 Wh, while daily AI workflows involving text, images, and short videos can reach 2–3 kWh . Training large frontier models can consume tens of gigawatt-hours, enough to power tens of thousands of homes for a year .

Future Projections

Global electricity consumption for data centers is projected to double to around 945 TWh by 2030, representing nearly 3% of total global electricity . In the U.S., AI-driven data center consumption could rise to 325–580 TWh by 2028, or 6.7–12% of national electricity generation . Growth is driven primarily by AI adoption, with the U.S. and China accounting for about 80% of the increase . Annual growth in data center electricity use is expected to be around 15% per year, far outpacing overall electricity demand .

Environmental Implications

High energy use from AI servers contributes to greenhouse gas emissions, water consumption for cooling, and electronic waste due to short hardware lifespans . Manufacturing GPUs and other high-performance components also requires rare earth minerals, adding to environmental impact . Without efficiency improvements, AI expansion could significantly strain power grids and exacerbate climate change.

Strategies for Efficiency

To mitigate energy consumption, the industry is focusing on:

• Optimizing AI models to reduce computational requirements.
• Developing greener infrastructure, including more efficient cooling and power systems.
• Software and hardware co-design to improve energy efficiency per operation.
• Scenario-based planning to anticipate energy demand and integrate renewable energy sources . In summary, AI servers are a rapidly growing driver of electricity demand, with significant environmental and operational implications. Efficient hardware, optimized models, and sustainable data center practices are critical to managing this energy footprint.