How Much Energy Do Data Centers Use? 15+ Data Center Energy Consumption Statistics for 2026

Data centers are quickly becoming one of the pillars of the modern economy, while also growing more and more controversial as the tech sector’s race toward Artificial General Intelligence (AGI) heats up.

These systems now consume enormous amounts of electricity and water, and it seems that we’re just witnessing a beginning of their true expansion, making data center energy usage one of the biggest infrastructure and utility challenges we’ve seen in a long time.

Key Data Center Energy Consumption Statistics for 2026 – Editor’s Choice

• The global data center sector accounted for 1.5% of all electricity use worldwide.
• Data center electricity use is expected to grow by around 15% annually through 2030.
• By 2028, US data center demand could rise to between 325 TWh and 580 TWh.
• The United States and China are expected to drive nearly 80% of global demand growth through 2030.
• Large data centers can consume up to 5 million gallons of water per day.

Global data centers consumed 448 TWh of electricity in 2025.

(Gartner)

Data centers are now responsible for more than 1.5% of all electricity consumption in the world, with projections showing an increase to 600 TWh in 2026.

The United States and China are expected to account for nearly 80% of data center electricity demand growth by 2030.

(IEA)

This means the global energy burden is highly concentrated in just a few regions. As a result, grid stress, utility bottlenecks, and infrastructure investment are becoming particularly important in those major technology markets.

Operational data center capacity now exceeds 55 GW globally.

(Cushman & Wakefield)

The sector is a global behemoth already, and is growing increasingly difficult for power systems around the world to ignore.

Global data center power demand is expected to surge 165% by 2030.

(Goldman Sachs Research)

The newest research from Goldman Sachs also estimates that AI is accelerating much faster than many power grids are designed to handle.

By 2028, data centers could account for 6.7% to 12% of US electricity demand.

(Belfer Center)

Server infrastructure and AI workloads are expected to dramatically impact the American power grid, like nothing we’ve seen before.

The average global data center PUE remained around 1.56 in 2025.

(ICIS)

PUE, or Power Usage Effectiveness, is one of the most common ways to measure how efficiently a data center uses energy. 

New facilities are designed to have a PUE of 1.44 or better, but the global average remains higher due to the high volume of older, less efficient legacy data centers.

Around 89% of North American data center capacity under construction was already precommitted by year-end 2025.

(JLL)

The latest numbers prove that tech firms are aggressively securing future capacity, as supply continues to highly outrun the current demand.

Active data center capacity in North America reached 39 GW by the end of 2025.

(JLL)

This is almost double the current global power capacity, signaling just how massive the buildout of data centers is right now in the US, with access to electricity becoming a larger issue with each passing day.

Large data centers can consume up to 5 million gallons of water per day.

(EESI)

While electricity gets most of the attention, water use is becoming a major issue as well. Cooling high-density server environments often requires substantial water resources, especially in hotter climates and larger facilities.

AI-related data centers globally could require up to 282 billion gallons (1,068 billion liters) of water annually by 2028.

(Morgan Stanley)

That figure underscores how quickly the resource footprint of AI infrastructure could expand over the next few years. It also broadens the conversation beyond electricity and into local water resilience, permitting, and community opposition.

AI-optimized servers are projected to account for 44% of all data center power consumption by 2030.

(Gartner)

This marks a massive shift from 2025, where AI servers represented only 21% of total demand.

The top five U.S. hyperscale companies are projected to invest $710 billion in capital expenditures in 2026 alone.

(JLL)

Much of this investment is dedicated to power-dense infrastructure and cooling technologies required to support the next generation of LLMs and generative AI tools.

64% of the 35 GW construction pipeline in North America is now located in "frontier markets" outside of traditional hubs.

(JLL)

As power becomes harder to secure in Northern Virginia and Silicon Valley, developers are moving to markets like Ohio and Texas where grid capacity is more accessible.

By the end of 2026, 50% of organizations will adopt "GreenOps" sustainability monitoring for their data center environments. 

(Gartner)

With the public and regulatory pressure mounting, half of all global enterprises are expected to implement real-time tracking of carbon footprints and water usage effectiveness (WUE) by the end of this year, representing a major jum from less than 10% in 2023.

The adoption of liquid cooling in AI data centers is projected to reach 40% in 2026.

(Accenture)

This transition is expected to provide energy savings for cooling of up to 50–60% compared to legacy air-based systems, but also comes with the rising public concern over water usage, especially in places like Arizona.