Between 2010 and 2020, something remarkable happened. Global data center energy consumption stayed flat, even as the amount of data we produced exploded. It seemed like we'd cracked the code: efficiency improvements were successfully counteracting the ever-increasing demands of the digital world. We saw data production soar from a mere 2 zettabytes to a staggering 64.2 zettabytes. How did we manage that? It was a combination of impressive efficiency gains in information and communications technology (ICT) and improvements in Power Usage Effectiveness (PUE).
However, around 2020, that trend shifted. Energy consumption started climbing again, mirroring the rise in data production. This change, as we explore in our whitepaper, Bending the Energy Curve, is largely due to the rise of power-hungry applications like AI and cloud computing. This raises some critical questions: Can the power grid handle this surge? What will be the impact on our transition to cleaner energy sources?
Graph showing total data center energy consumption. A clear trend break is seen in 2020 when the energy curve shifts from flat to increasing.
Our research suggests that even seemingly modest improvements in data center efficiency can have a huge impact on overall energy consumption, even with data production continuing to skyrocket. To understand this relationship between performance improvements and energy use, we developed a metric we call the ICT Performance Factor. It's calculated by dividing the annual ICT energy consumption (in terawatt-hours) by the amount of data produced (in zettabytes).
The paper also examines the potential of further PUE improvements. PUE, remember, is a measure of how efficiently a data center uses energy. A PUE of 1.0 is perfect – all energy goes to computing. A PUE of 2.0 means half the energy is wasted. The average PUE has been steadily declining, from around 2.0 in 2010 to 1.51 in 2024, and is projected to reach 1.43 by 2030.
But we believe we can be more ambitious. For new data centers, we're targeting a Power Usage Effectiveness (PUE) of 1.2, and for some existing ones, 1.3, by 2030. Reaching this goal could lower global data center energy use by 3.6%, saving a large amount of energy (38 terawatt-hours).
Beyond PUE, we see even bigger savings with ICT efficiencies. By improving how efficiently data centers use their computing power. We think that with focused effort in areas like AI-optimized hardware and better software, we can significantly reduce the energy needed for data processing by 2030.
Results of modest acceleration to infrastructure and ICT improvements shows the impact on "bending" the growth curve and showing that, once again, it is possible to decouple data center energy growth from digitalization drivers
By focusing on efficiency, the industry can concentrate on securing sustainable energy sources rather than struggling with an unmanageable surge in energy demand.
We identified several factors contributing to the break in the previously flat energy consumption curve around 2020:
01
Limitations in shifting workloads to more efficient cloud environments.
02
Diminishing CPU performance gains.
03
Pandemic-related delays in silicon releases.
Our paper emphasizes four key steps for data center operators and industry stakeholders:
Next Steps
01
Set Ambitious PUE Targets
Aim for a PUE of 1.2 or lower for all new data center builds.
02
Shift to More Efficient Computing Methods
Embrace accelerated computing platforms, optimize workload configurations, and use intelligent scheduling algorithms.
03
Invest in Sustainable Energy Solutions
Explore behind-the-meter power generation options.
04
Foster Industry Collaboration
A unified approach is essential to decouple data growth from energy consumption.
The core message is this: While the projected surge in data center energy consumption is a valid concern, it's not an insurmountable challenge. By prioritizing efficiency, innovation, and collaboration, the industry can navigate this energy transition and create a truly sustainable digital future. The challenges are real, but so are the opportunities. By thinking ahead and working together, we can harness the power of digitalization while ensuring a sustainable future for all.
Between 2010 and 2020, something remarkable happened. Global data center energy consumption stayed flat, even as the amount of data we produced exploded. It seemed like we'd cracked the code: efficiency improvements were successfully counteracting the ever-increasing demands of the digital world. We saw data production soar from a mere 2 zettabytes to a staggering 64.2 zettabytes. How did we manage that? It was a combination of impressive efficiency gains in information and communications technology (ICT) and improvements in Power Usage Effectiveness (PUE).
However, around 2020, that trend shifted. Energy consumption started climbing again, mirroring the rise in data production. This change, as we explore in our whitepaper, Bending the Energy Curve, is largely due to the rise of power-hungry applications like AI and cloud computing. This raises some critical questions: Can the power grid handle this surge? What will be the impact on our transition to cleaner energy sources?
Graph showing total data center energy consumption. A clear trend break is seen in 2020 when the energy curve shifts from flat to increasing.
Our research suggests that even seemingly modest improvements in data center efficiency can have a huge impact on overall energy consumption, even with data production continuing to skyrocket. To understand this relationship between performance improvements and energy use, we developed a metric we call the ICT Performance Factor. It's calculated by dividing the annual ICT energy consumption (in terawatt-hours) by the amount of data produced (in zettabytes).
The paper also examines the potential of further PUE improvements. PUE, remember, is a measure of how efficiently a data center uses energy. A PUE of 1.0 is perfect – all energy goes to computing. A PUE of 2.0 means half the energy is wasted. The average PUE has been steadily declining, from around 2.0 in 2010 to 1.51 in 2024, and is projected to reach 1.43 by 2030.
But we believe we can be more ambitious. For new data centers, we're targeting a Power Usage Effectiveness (PUE) of 1.2, and for some existing ones, 1.3, by 2030. Reaching this goal could lower global data center energy use by 3.6%, saving a large amount of energy (38 terawatt-hours).
Beyond PUE, we see even bigger savings with ICT efficiencies. By improving how efficiently data centers use their computing power. We think that with focused effort in areas like AI-optimized hardware and better software, we can significantly reduce the energy needed for data processing by 2030.
Results of modest acceleration to infrastructure and ICT improvements shows the impact on "bending" the growth curve and showing that, once again, it is possible to decouple data center energy growth from digitalization drivers
By focusing on efficiency, the industry can concentrate on securing sustainable energy sources rather than struggling with an unmanageable surge in energy demand.
We identified several factors contributing to the break in the previously flat energy consumption curve around 2020:
01
Limitations in shifting workloads to more efficient cloud environments.
02
Diminishing CPU performance gains.
03
Pandemic-related delays in silicon releases.
Our paper emphasizes four key steps for data center operators and industry stakeholders:
Next Steps
01
Set Ambitious PUE Targets
Aim for a PUE of 1.2 or lower for all new data center builds.
02
Shift to More Efficient Computing Methods
Embrace accelerated computing platforms, optimize workload configurations, and use intelligent scheduling algorithms.
03
Invest in Sustainable Energy Solutions
Explore behind-the-meter power generation options.
04
Foster Industry Collaboration
A unified approach is essential to decouple data growth from energy consumption.
The core message is this: While the projected surge in data center energy consumption is a valid concern, it's not an insurmountable challenge. By prioritizing efficiency, innovation, and collaboration, the industry can navigate this energy transition and create a truly sustainable digital future. The challenges are real, but so are the opportunities. By thinking ahead and working together, we can harness the power of digitalization while ensuring a sustainable future for all.
Want to see the full analysis, including the three-phase trend, specific technology recommendations, and the potential for a 25.5% reduction in grid energy burden? Download the full report!
