How much will electricity consumption from data centres grow from today’s 1% of the global total? 40-fold by 2030? Or a more manageable 5-fold? Or less? Sean Ratka and Francisco Boshell at IRENA try to answer this question by looking at the innovations being made by the tech industry to drive down power costs and emissions. The evidence is promising. Though data centre computing output jumped 6- fold between 2010 and 2018, their energy consumption rose only 6%. Big tech love efficiency. The authors draw attention to the shift to centralised cloud computing, the use of artificial intelligence to optimise consumption patterns, locating servers in cold climates (even underwater!), and buying power from renewables as their costs continue to drop. Demand-side management (non-critical numbers can be crunched and spat out at any time) and sector coupling (selling the waste heat) also add to the efficiency gains. All the big tech giants are doing it: Amazon, Google, Facebook, Microsoft. The authors recommend that this wide variety of solutions be scrutinised by other sectors struggling to decarbonise.
Energy consumption from data centres is currently attracting more attention as a matter of concern, as global computing capacity continues to rapidly increase. But rather than seeing them as a threat to the sustainable energy transition, and the wider climate goals outlined in the Paris Agreement, data centres may actually present an opportunity to accelerate the transition.
IRENA showed in its Innovation Landscape for the Power Sector study, that data centres sit at the nexus of energy efficiency, renewable energy, and burgeoning data economy enabled by digitalisation. By integrating the latest technologies and leveraging the increasingly attractive economics of renewables and the increased efficiencies made possible by artificial intelligence, data centres are paving a path that other power-hungry industries could follow.
Current and projected energy consumption from data centres
Due to the exponential growth in data collection and use, the need for data centres continues to grow. The servers, storage equipment, backups, and power cooling infrastructure in data centres require electricity, and lots of it. According to Google, each internet search uses about 0.0003 kWh, or 1 KJ, of energy (Google, 2009). To put this in perspective, 0.0003 kWh could power a 60W light bulb for about 17 seconds. While this may seem low, looking at the current massive flow of data in a more digitalised world, the impacts on energy use and carbon emissions from data centres quickly add up.
Globally, data centres accounted for approximately 1% (or 205 TWh) of global electricity use in 2018 (Masanet et al., 2020; Pearce, Fred, 2018) and emit as much CO2 as the commercial airline industry (Data Economy, 2017). Estimates suggest that annual electricity demand from data centres could grow to as much as 8,000 TWh by 2030 under worst case scenarios and to as low as 1,100 TWh under the best-case scenarios (Nature, 2018). Based on recent energy consumption growth rates and efficiency gains, best case or even lower electricity consumption scenarios seem likely, as we discuss next.
Despite concerns raised surrounding the growth of data centres (and the electricity required to power and cool them), a new study of data centres globally found that while their computing output jumped six fold from 2010 to 2018, their energy consumption rose only 6 percent (Masanet et al, 2020). Essentially, massive efficiency gains have allowed computing power to sharply increase while power consumption remained largely flat.
These massive efficiency gains have mainly come from processor efficiency improvements, reductions in idle power, increased storage drive density and slowing server growth. The shift to cloud computing which relies on hyperscale data centres, the largest and most efficient type of data centre, has further accelerated efficiency improvements. These hyperscale data centres, run by the likes of Amazon Web Services (AWS), Google, Facebook, and others, usually boast the most cutting edge technology (facility designs, cooling systems, and workload-optimised equipment) in order to reduce (energy) costs.
In fact, on average they only require 16% of the power as compared to on-premises infrastructure for the same amount of computations (AWS, 2020). Because of this, moving on-premise workloads to large-scale data centres can lower the workload carbon footprint by 88% for the median surveyed US enterprise data centres (451 Research, 2019).
Artificial intelligence (AI) is playing a crucial role as well. Google’s DeepMind AI, for example, reduced the energy used for cooling at one of Google’s data centres by 40% in 2016 (a 15% overall reduction in power usage) using only historical data collected from sensors within the data centre (e.g. temperatures, power, pump speeds, setpoints) to improve data centre energy efficiency. The AI system predicts the future temperature and pressure of the data centre over the next hour and gives recommendations to turn the consumption on or off.
Besides technological innovations, efficiency gains have also been achieved by locating data centres in cold regions, where servers can be cooled using outside air or water, thereby reducing energy usage. Finland, for example, boasts hyperscale data centres with some of the lowest energy-related costs globally, due to its cool climate. The market for “hyperscale” data centres in Finland was valued at $5 billion in 2018 and is expected to be worth $17 billion by 2025 (Wintergreen, 2019).
Microsoft is also experimenting with innovative technologies such as installing data centres on the seabed, in order to benefit from the passive cooling provided by the cold seawater as well as the abundant marine energy provided by tides and waves to generate the electricity needed to power servers. Microsoft’s “Project Natick” is one such project currently being tested off the coast of Scotland and includes 12 racks containing 864 servers packed in a 40-foot long container on the seabed (Microsoft, 2018).
Cheaper renewable energy is powering data centres
The rapid deployment of renewable power generation technologies globally, combined with high learning rates, has driven down costs. This trend is projected to continue, making renewables increasingly competitive with fossil fuels in countries across the world and the least-cost option in a growing number of markets. This bodes well for data centres, as electricity accounts for as much as 70% of their total operating costs (SSG, 2017).
IRENA’s latest report shows that newly installed renewable power capacity increasingly costs less than the cheapest power generation options based on fossil fuels. More than half of the renewable capacity added in 2019 achieved lower electricity costs than new coal. New solar and wind projects are also undercutting the cheapest existing coal-fired plants, the report finds.
As the demand for cloud computing services has increased, the share of renewables powering these data centres has increased. Electricity systems themselves have become more complex over the past few decades with increased decentralisation, decarbonisation and digitalisation, leading to increased reliance on data analysis. In effect, renewables are contributing to the increased demand for data centres, while simultaneously helping to decarbonise them.
In many ways cloud computing companies are leading the way in terms of corporate sustainability, driven by the increasingly attractive economic argument of renewables. In many regions, renewables are simply the cheapest option, and data centres tend to congregate where power is cheap. For example, when considering Bitcoin mining, which also relies heavily on electricity to run mining equipment and cooling infrastructure, these facilities are largely located in areas with cool climates and an abundance of cheap renewable energy (mainly hydropower and wind), including China’s Sichuan province, Iceland, Quebec, British Columbia, Norway, and others.
Tech firms like renewables
The commitment of technology firms to renewables is evidenced in the amount of renewable energy they buy. The Renewable Energy Buyer’s Alliance ranked the top 10 buyers of renewable energy in the United States in 2019, with many technology companies making the list.
AWS, Google, Facebook, and Microsoft have all recently announced large-scale renewable energy purchases. Google announced in 2017 that it achieved 100% renewable energy across all of its operations, including data centres. Importantly, Google, and the rest of the major players in this space rely on renewable energy credits (RECs) to offset their fossil fuel usage and claim 100% renewable use while still being connected to grids that rely on fossil fuels for power generation.
AWS, by far the biggest cloud computing provider, with well over one-third of the market has a long-term commitment to achieve 100% renewable energy usage and reached 50% across the entire Amazon company in 2018, when RECs are factored in. Microsoft has been carbon neutral since 2012, and including its RECs, it has run on 100 percent renewable energy since 2014 (Wired, 2019). In fact, Microsoft aims to be carbon negative by 2030, and by 2050 remove from the environment all the carbon the company has emitted either directly or by electrical consumption since it was founded in 1975 (Microsoft, 2020).
Data centres can offer flexible demand
As sources of flexibility are more valued and compensated in power markets, engaging in demand-side management can further decrease the cost of ownership of a data centre. Given their vast energy usage and their highly automated and “intelligent” nature (due to the integration of IoT devices and machine learning algorithms to optimise power usage), data centres provide a great opportunity to add to the flexibility of grids through demand-side management. This increased flexibility would allow grids to integrate higher shares of variable sources of energy, such as solar and wind. By either ramping up or down their power usage based on price signals, data centres could reduce peak power demand during times with limited solar and wind generation and conversely, consume more power during times of excess renewables generation. (Basmadjian, 2019).
Google has also designed and deployed an innovative system for their hyperscale data centres to shift the timing of many compute tasks to when low-carbon power sources, like wind and solar, are most plentiful. According to Google, this is done without additional computer hardware and without affecting Google’s services. Shifting the timing of non-urgent compute tasks—like creating new filter features on Google Photos, YouTube video processing, or adding new words to Google Translate—helps reduce the electricity grid’s carbon footprint by relying more heavily on variable renewable generation, such as wind and solar.
Sector coupling: using waste heat from data centres
Sector coupling is seen as a way to shape a more integrated renewable energy system by interconnecting energy consuming sectors. Data centres may have a potential role to play, by using renewable electricity for their operations and subsequently using the waste heat (a natural by-product of servers) to heat large amounts of water for heating nearby houses or businesses, thus indirectly electrifying with renewables the heating sector.
Facebook’s data centre in Odense, Denmark, for example, captures excess heat generated by their servers and recycles it to provide heat to the local community. The heat pump (power-to-heat) is also supported by 100% renewable energy. This energy will then be directed into the local district heating system, operated by district heating company Fjernvarme Fyn (Facebook, 2019). If data centres source their electricity from renewable sources, not only can they increase the sustainability of their own processes, but of their surrounding areas as well.
Role model for other energy intensive sectors
Technology giants are some of the largest consumers of electricity. But recent developments discussed in this article indicate that instead of acting as a roadblock, they are potentially a role model for the energy transition and are some of the most innovative companies when it comes to pioneering new strategies for leveraging renewables while increasing efficiencies. Strategies outlined above, include:
- locating data centres in cold locations with abundant and cheap renewable electricity;
- enabling data centres to provide services in power flexibility markets;
- coupling energy streams from different applications; and
- use of digital technologies to optimise operations;
- heavily investing in energy efficiency should be replicated by other sectors.
Those strategies have been largely driven by economics and consumer sentiment, pushing highly-visible companies to be more “green”.
Doomsday forecasts which base future energy demand projections on current demand in data centres often fail to take into account the innovations which continue to reduce energy demand in the information technology sector. If efficiencies continue to increase, alongside the growing reliance on renewable energy, our increasing need for computing power and data analysis may not lead to environmental destruction, but rather accelerate the overall sustainable energy transition by supporting renewable energy producers and accelerating the development of innovative technologies. These large energy consumers are showing us what is possible when economics, consumer pressure to be green, and innovations in digitalisation converge.
AWS (2020), “AWS & Sustainability”, https://aws.amazon.com/about-aws/sustainability/
Basmadjian, R. (2019), “Flexibility-Based Energy and Demand Management in
Data Centers: A Case Study for Cloud Computing” https://www.mdpi.com/1996-1073/12/17/3301/pdf
Data Economy (2017), “Data Centers ‘Going Green’ To Reduce A Carbon Footprint Larger Than The Airline Industry”, https://data-economy.com/data-centers-going-green-to-reduce-a-carbon-footprint-larger-than-the-airline-industry/
Evans, R., and J. Gao (2016), “DeepMind AI reduces Google data centre cooling bill by 40%”, Google DeepMind, https://deepmind.com/blog/deepmind-ai-reduces-google-data-centrecooling-bill-40
Facebook (2019), “Denmark data center to warm local community”, https://sustainability.fb.com/wp-content/uploads/2019/06/Waste_Heat_Recovery_Final_Jun2019.pdf
Google (2009), “Powering a Google search”, https://googleblog.blogspot.com/2009/01/powering-google-search.html
Google, 2019. “Our biggest renewable energy purchase ever” https://blog.google/outreach-initiatives/sustainability/our-biggest-renewable-energy-purchase-ever/
IRENA (2020), Renewable Power Generation Costs in 2019, International Renewable Energy Agency, Abu Dhabi. https://irena.org/publications/2020/Jun/Renewable-Power-Costs-in-2019
IRENA (2019), Innovation landscape brief: Artificial intelligence and big data, International Renewable Energy Agency, Abu Dhabi. https://www.irena.org/publications/2019/Sep/Artificial-Intelligence-and-Big-Data
Masanet, E., Shehabi, N., Lei, N., Smith, S., Koomey, J. (2020), “Recalibrating global data center energy-use estimates” DOI: 10.1126/science.aba3758, Science 367 (6481), 984-986.
Microsoft (2020), “Microsoft will be carbon negative by 2030”, https://blogs.microsoft.com/blog/2020/01/16/microsoft-will-be-carbon-negative-by-2030/
Microsoft (2018), “Under the sea, Microsoft tests a datacenter that’s quick to deploy, could provide internet connectivity for years”, https://news.microsoft.com/features/under-the-sea-microsoft-tests-a-datacenter-thats-quick-to-deploy-could-provide-internet-connectivity-for-years/
Nature (2018), “How to stop data centres from gobbling up the world’s electricity”, https://www.nature.com/articles/d41586-018-06610-y
Pearce, Fred (2018), “ Energy Hogs: Can World’s Huge Data Centers Be Made More Efficient?”, https://e360.yale.edu/features/energy-hogs-can-huge-data-centers-be-made-more-efficient
REBA (2020), “Renewable Energy Buyers Alliance Announces Top 10 U.S. Large Energy Buyers in 2019”, https://rebuyers.org/blog/renewable-energy-buyers-alliance-announces-top-10-u-s-large-energy-buyers-in-2019/
SSG (2017), “Power in the Data Center and its Cost Across the U.S.”, Site Selection Group. https://info.siteselectiongroup.com/blog/power-in-the-data-center-and-its-costs-across-the-united-states
Wintergreen (2019), “The Hyperscale Data Center Market Boom in Finland 2019-2025 – Market Expected to Reach $17 Billion by 2025, Growing from $5 Billion in 2018”, https://www.globenewswire.com/news-release/2019/11/14/1947059/0/en/The-Hyperscale-Data-Center-Market-Boom-in-Finland-2019-2025-Market-Expected-to-Reach-17-Billion-by-2025-Growing-from-5-Billion-in-2018.html
Wired (2019), “Amazon, Google, Microsoft: Here’s Who Has the Greenest Cloud”, https://www.wired.com/story/amazon-google-microsoft-green-clouds-and-hyperscale-data-centers/
- In the United States ↑