United States Data Center Energy Usage Report

About This Document

This Lawrence Berkeley National Laboratory study, supported by the U.S. Department of Energy’s Federal Energy Management Program, estimates energy use by U.S. data centers from 2000 to 2020. The model covers servers, storage systems, networking equipment, and facility infrastructure, using data on the installed base of IT equipment and PUE by facility type. The report compares the current trend with a counterfactual scenario in which 2010 efficiency levels are maintained, and examines the potential of improved management, best practices, and workload migration to hyperscale data centers.

Key Takeaways

• In 2014, U.S. data centers consumed approximately 70 billion kWh, or 1.8% of U.S. electricity use; under current trends, consumption in 2020 was projected to be approximately 73 billion kWh.
• Growth in energy consumption slowed sharply: approximately 4% from 2010 to 2014, compared with nearly 90% from 2000 to 2005, despite continued growth in digital demand.
• The main drivers of the slowdown were slower growth in the server installed base; virtualization and consolidation; improved energy efficiency of servers, storage, networks, and facility infrastructure; and growth in the hyperscale segment.
• If key efficiency parameters had remained at 2010 levels, 2014 energy consumption would have been 60% higher than the actual estimate, and 2020 consumption would have been 170% higher than under the Current Trends scenario.
• The Improved Management, Best Practices, and Hyperscale Shift scenarios indicate substantial additional potential: up to 33 billion kWh in annual savings by 2020, equivalent to 45% relative to current trends.

Key Figures

U.S. data center energy consumption

70 billion kWh

Estimate for 2014, including servers, storage, networking, and supporting infrastructure.

Data centers' share of U.S. electricity consumption

1,8 %

2014.

Forecast U.S. data center energy consumption

73 billion kWh

Around 2020 under the Current Trends scenario.

Growth in data center energy consumption

4 %

Estimated growth in 2010–2014.

Forecast growth in data center energy consumption

4 %

Forecast for 2014–2020 under the Current Trends scenario.

Growth in data center energy consumption

24 %

2005–2010.

Growth in data center energy consumption

almost 90 %

2000–2005.

Weighted-average PUE

1,89

Value recorded at the 2010 level in the counterfactual scenario.

PUE of large hyperscale data centers

1,1 or lower

Achievable level according to the report; small data centers may have a PUE above 2.0.

Excess energy consumption without further efficiency improvements

60 %

2014 energy consumption relative to the current estimate, if key metrics had remained at 2010 levels.

Excess energy consumption without further efficiency improvements

170 %

2020 forecast relative to the Current Trends scenario.

Electricity savings

100 billion kWh

Estimated industry-wide savings in 2010–2014 resulting from the continuation of energy-efficiency practices.

Additional electricity savings

520 billion kWh

Estimate for 2015–2020 assuming the continuation of Current Trends practices.

Total electricity savings

620 billion kWh

Estimate for 2010–2020.

Potential annual savings

up to 33 billion kWh/year

By 2020 under the Improved Management, Best Practices, and Hyperscale Shift scenarios.

Potential reduction relative to Current Trends

45 %

By 2020 for combined efficiency-improvement scenarios.

Share of servers in hyperscale data centers

40 %

Modeled share in 2020; up from 0% in 2005.

Average service life of servers and storage systems

4,4 years

Assumed value for calculating the installed base.

Average server utilization

14 %

2010 industry-average level used in the counterfactual scenario.

Size of an individual hyperscale data center module

50 000 sq. ft.

Average module size; a facility may include up to 8 modules.

Practical Value for Data Center Owners

For a data center owner or project team, the report provides a historical benchmark for energy-use structure and shows that assessing PUE alone is insufficient: the greatest impact comes from coordinated work with the IT load, including consolidation, removal of inactive servers, higher utilization, and migration of suitable workloads to large, efficient platforms. When developing a modernization program, it is useful to compare PUE and the IT-load profile separately across facility types. Small server rooms have limited potential for infrastructure improvements, whereas a shift to centralized or hyperscale architectures can simultaneously reduce excess server capacity and cooling and power-distribution overheads.

Where It Applies

Topics