Data centers are at the heart of ICT[1], but for many the inner workings and the wider impact are relatively unknown. However, what goes on below the cloud or in private data center must be of interest to us all, as these energy sapping IT workhorses might soon require more energy as all of Japan consumes every year.
The Rise Of The Data Center
Mainly driven by AI, cloud, cryptocurrency and the fact that IT is now the business of every business, the number of data centers will be increasing. Back in 2020 there were about 8,000 data centers in the world[2], consuming about 340 TWh[3]. The prediction is that he number of data centers might reach 15,000 – 17,500 by 2030, potentially consuming maybe up to 1,000 TWh – this is what Japan consumes per year; if this prediction becomes real, and if all data centers were a single country, it would be the 5th highest ranked country energy consumer in the world, just behind Russia, India, USA and China.
As with any prediction they might not come true, as new innovations allow for a reduction in energy consumption, whilst increasing compute performance. Innovation in areas like artificial intelligence, that focuses on power reductions or further development in the chip design, like advanced processor architectures or neuromorphic compute, might allow for lowering energy footprint with whilst increasing performance characteristics.
However, the big challenge remains: how to increase the number of data centers whilst minimising the potential environmental impact?
What Is A Data Center?
A data center is a physical facility that organizations use to house their critical applications and data. It is a dedicated space within a building, or a group of buildings, used to house computer systems and associated components, such as telecommunications and storage systems. Even though data centers have been around since the early 1950 there is no agreed definition, making it difficult to establish precise numbers.
What Is The Issue With The Increase Of Data Centers?
Among the most pressing issues related to the rise in data centers are supply chain constraints, community concerns over location and resource use, and most pressing, the growing environmental impact and the power infrastructure demand.
The Growing Environmental Impact
The environmental impact of a data center depends on many factors, like the impact of constructing the facilities in the first place, the power usage (see below), growing e-waste disposal, inefficient water usage, use of diesel generators and large batteries to cater for power outages, and the creation of micro-climates by cooling systems.
The way a data center is being operated, and the way it has been designed are critical. Some data centers repy fully on air-conditioning to cool all equipment’s, which in turn can consume significant power and water.
Another factor is size. With the advent of cloud, the size of data centers has significantly increased. Large data centers, often referred to as hyperscale data centers, can cover millions of square feet and contain tens of thousands of servers. For instance, The Citadel located in Reno, Nevada, has a power capacity of 650 megawatts and covers up to 7.2 million square feet (670,000 square meters). The Range International Information Group operates a data center in Langfang, China, this data center that covers 6.6 million square feet (610,000 square meters) – that’s about 85 football pitches. Next to the build-and-operate impact, these hyperscale data centers require significant backup equipments like diesel generators and large batteries, which in turn adds to the environmental impact.
Power Infrastructure demand
The only way seems to be up. Across the globe the demand for power, driven by the increase in data centers is applying significant pressure onto the electricity networks.
The International Energy Agency stated last year that electricity consumption from data centers, artificial intelligence (AI) and the cryptocurrency sector could double by 2026. Furthermore, the IEA estimated that almost 2% of total global electricity demand is driven by data centers.
An EU paper issued at the end of 2022, stated that data centers accounted for 2.7% of EU electricity demand in 2018 and was expected to rise by 200% by 2030. The IEA has forecast that by 2026 data centers could account for 29% of new electricity demand in Europe. The electricity demand of data centers in Europe is expected to be at 3.2 percent of the EU total by 2030. To put this into context, the IEA estimates that Ireland could see an almost 2-fold and Denmark a 6-fold increase in the electricity use by data centers by 2030.
In the US, which has by far the highest concentration of data centers in the world (about 33% of all data centers), the demand might lead to an increase of CO2 from 40 million tons to 67 million tons, as most of the demand will have to be serviced in the short-term by high-carbon power stations, like gas or coal. In total, all data centers consume ~4% of the US electricity supply, increasing to 6% of total electricity demand by 2026.
As noted in the introduction section of this article these predications might be inaccurate, as it is difficult to project the innovations and future developments that might reverse the trend. However, it is hard to see that, with the increase in compute demand and the related rise of data centers the energy consumption would stay flat or reduces compared with today.
Where Is The Demand Being Generated?
Electricity demand in data centers is mainly from two aspects, the compute (so servers running) and cooling (mainly from air conditioning). Computing accounts for 40% of electricity demand, whereas cooling requirements to achieve stable processing efficiency makes up about another 40%. The remaining 20% comes from other associated IT equipment.
How To Measure The Energy Efficiency Of A Data Center?
Several measures are available to assess the efficiency of a data center. Energy Reuse Effectiveness (ERE), Carbon Usage Effectiveness (CUE) or Data Center Infrastructure Efficiency (DCiE). Most widely Power Usage Effectiveness (PUE) is being used. PUE is defined as the ratio of the total annual energy to run the data center facility to the total annual energy drawn by all IT equipment. Today, a PUE rating of 1.6 is standard, with 1.4 being good and 1.1 excellent. A PUE (Power Usage Effectiveness) of 1.6 means that for every 1 watt of power used by the IT equipment in the data center, 0.6 watts are used for cooling, power distribution, and other non-IT infrastructure. Essentially, the data center is using 60% more energy than what is consumed by the IT equipment alone. So, the lower the number the better. A 1.1 would only lose 10% of energy compared with 60% for a PUE of 1.6.
What Can We Do To Advert The Potential CO2 Impact?
There are five main ways: 1) Regulations, 2) Using renewable or low-carbon energy supplies, 3) Building energy efficient data centers, 4) Reducing the compute demand, 5) Considering sustainability as a non-functional requirement when designing a solution and 6) closely monitoring all sustainability aspects.
- Regulation
One way to reverse the trend of data centers producing more CO2 is via regulations. For instance, last year the EU has adopted a new delegated regulation on the first phase for establishing an EU-wide scheme to rate the sustainability of EU data centers. Under the new, recast Energy Efficiency Directive, this legislation, which came in force in 2024, requires data center operators to report the key performance indicators every year.
- Using Renewable or Low-Carbon Energy Supplies
Another route is shifting towards renewable or low-carbon energy supplies. The European Commission has set ambitious renewable energy targets, including a proposal to increase the share of renewable energy sources in the EU’s overall energy mix to at least 42.5% by 2030.
- Building Energy Efficient Data Centers
Building much more energy efficient data centers is another route. Fifteen years ago, I served as the Chief Architect for a significant IT transformation program for a UK Government Department, delivering what was then the most sustainable IT initiative. Driven by cost savings, future proofing and long-lasting sustainability, one of the main components was a futuristic data center which re-used an existing warehouse, applied a modular construction relying on 95% recycled materials, made use of flywheels instead of environmentally damaging battery installations, fresh air cooling instead of energy and water intensive air conditioning units and applied a fully service based, unattended and automated operations. In controlled factory testing, the data center modules achieve a PUE of 1.08. Under operational conditions, adding in electrical losses, we achieved a data center PUE of 1.1. What was futuristic 15 years ago is now state-of-the-art.
- Reducing the Compute Demand
Caused by over-provisioning of capacity, data centers often operate at low utilizations, with an industry average of 10%-50% utilization, potentially wasting valuable energy, without providing much value. Careful forecasting and effective resource allocation can assist with reducing the waste, thus maximising the energy usage. Often data centers host latency intensive workload that is critical to the business, and balancing latency-critical scale-out workloads as energy efficient as possible without degrading the performance characteristics can be very complex. For example, Google released their toolset as open source – OLDIsim, or OnLine Data Intensive simulator.
Another way of reducing the compute demand is avoiding building something new and instead reusing an existing solution. This means being more sensible with new compute demand, by considering if the requirement can be met differently.
Next to minimising overprovisioning and considering maybe re-using a solution, there are further ways of reducing the compute demand. For example, why building, operating and maintaining a separate development, test, training user acceptance test and a live environment? Assuming the applications design supports this, only use dev and live environment. Or why keeping your server running during the night if 24/7 is not needed?
- Consider Sustainability as a Non-functional Requirement
Also, make sure that when designing a solution, you consider sustainability as a non-functional requirement right at the start of the solution creation process, instead of in the end, ensuring that there is a balance between value for money, agility, compliance and sustainability throughout.
- Closely Monitor Sustainability Aspects
To reduce an existing or minimize a future planned sustainability impact of a data center service, a current (or predicated) baseline is needed. However, measuring the current or predicting a future sustainability impact like CO2 footprint of all direct or indirect assets that draws energy can be difficult. Over the past couple of years various tools and methods have been established to ascertain and report on sustainability aspects. For instance, at Capgemini we developed a state-of-the-art Energy Command Center, helping us to reduce our own facility energy consumption by 29%.
The following areas are key:
- Identification of energy and CO2 optimization measures
- Improved data visibility for more informed decision-making
- Ongoing improvements in sustainability practices with reduced manual workload
- Strengthened compliance with emerging regulations
Summary
The global increase in data centers from 8,000 five years ago to potentially 17,000 in the next five years seems inevitable. However, the environmental impact does not have to be inevitable. By implementing regulations, using cleaner electricity, constructing more efficient data centers, being more mindful of our compute power usage, putting sustainability at the center of any solution design, and closely monitoring all sustainability related aspects, we can safely increase the number of data centers whilst reducing their overall environmental impact.
References
- Capgemini TechnoVision 2025, https://www.capgemini.com/wp-content/uploads/2025/02/TV25-Report_Web-version.pdf
- United States International Trade Commission, https://www.usitc.gov/publications/332/executive_briefings/ebot_data_centers_around_the_world.pdf
- Wiki, https://en.wikipedia.org/wiki/List_of_countries_by_electricity_consumption
- Wiki, https://en.wikipedia.org/wiki/Data_center
- Switch, https://www.switch.com/tahoe-reno/
- Webopedia, https://www.webopedia.com/technology/10-biggest-data-centers-in-the-world/
- International Energy Agency, https://iea.blob.core.windows.net/assets/6b2fd954-2017-408e-bf08-952fdd62118a/Electricity2024-Analysisandforecastto2026.pdf
- European Commission, https://ec.europa.eu/commission/presscorner/api/files/document/print/en/qanda_22_6229/QANDA_22_6229_EN.pdf
- International Energy Agency, https://iea.blob.core.windows.net/assets/6b2fd954-2017-408e-bf08-952fdd62118a/Electricity2024-Analysisandforecastto2026.pdf
- European Commission, https://eur-lex.europa.eu/resource.html?uri=cellar:12e835e2-81af-11eb-9ac9-01aa75ed71a1.0001.02/DOC_1&format=PDF
- European Commission, https://energy.ec.europa.eu/news/commission-adopts-eu-wide-scheme-rating-sustainability-data-centers-2024-03-15_en
- United States International Trade Commission, https://www.usitc.gov/publications/332/executive_briefings/ebot_data_centers_around_the_world.pdf?utm_source=chatgpt.com
- SPGlobal, https://www.spglobal.com/ratings/en/research/articles/241030-data-centers-rapid-growth-creates-opportunities-and-issues-13307638
- United States International Trade Commission, https://iea.blob.core.windows.net/assets/6b2fd954-2017-408e-bf08-952fdd62118a/Electricity2024-Analysisandforecastto2026.pdf
- United States International Trade Commission, https://www.iea-4e.org/wp-content/uploads/2022/10/EDNA-Studies-Metrics-for-data-center-efficiency-Final.pdf
- US Department of Energy, https://www.energy.gov/sites/default/files/2024-07/best-practice-guide-data-center-design.pdf#page=38&zoom=100,93,316
- European Commission, https://ec.europa.eu/info/law/better-regulation/have-your-say/initiatives/13818-Data-centers-in-Europe-reporting-scheme_en
- European Union, https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=OJ%3AJOL_2023_231_R_0001&qid=1695186598766
- European Commission, https://energy.ec.europa.eu/topics/renewable-energy/renewable-energy-directive-targets-and-rules/renewable-energy-targets_en
- US Department of Energy, https://www.energy.gov/sites/default/files/2024-07/best-practice-guide-data-center-design.pdf#page=38&zoom=100,93,316
- Stanford University, https://csl.stanford.edu/~christos/publications/2015.david_lo.phd_thesis.pdf
- Gunnar Menzel, https://www.linkedin.com/pulse/architecting-sustainable-solutions-gunnar-menzel
- Gunnar Menzel, https://www.linkedin.com/pulse/calculating-your-servers-co2-footprint-gunnar-menzel
- Capgemini, https://www.youtube.com/watch?v=2sY80SfRhaI
- Capgemini, https://www.capgemini.com/in-en/solutions/capgemini-energy-command-center/
- Capgemini, https://www.capgemini.com/de-de/wp-content/uploads/sites/8/2025/01/Siemens-Corporate-IT-sustainability-Client-Story.pdf
[1] Information and Communication Technology.
[2] Exact numbers are difficult to obtain as the difference between server-room and data center are ambiguous
[3] TWh = A terawatt-hour (TWh) is a unit of energy that represents one trillion watt-hours. 340 TWh is Turkey’s annual electricity consumption