Home
hamburger menu icon

Information Technology

As information technology becomes more central to the world economy and society, its energy demands will grow rapidly.

Data Centers and Artificial Intelligence

In the 2020s, artificial intelligence has been the leading technological revolution driving change in the economy. Although data centers, the growth of which is driven by the training and usage of AI models, are a small part of the world energy mix, they are growing rapidly. The following worldwide data center electricity demand has been observed and projected by the International Energy Agency as of 2025 1.

Due to great uncertainty about the trajectory of AI development, as well as broader macroeconomic trends, the IEA projects data center electricity demand in 2035 as ranging from 700 TWh to 1700 TWh. By comparison, world electricity consumption was just under 30,000 TWh in 2023 2. The rate of growth of data center electricity demand has been 12% per year from 2017 to 2024, more than four times as much as the rate of growth of electricity demand overall 1. Data centers will account for 10% of world electricity demand growth to 2030; comparable figures are 5% in low-income countries, where other demand growths are also strong, 20% in wealthy countries, and nearly 50% in the United States.

Estimating current data center energy usage, let alone projecting into the future, has been a challenge.

Kamiya and Coroamă 3 review several studies of data center energy usage since 2014. They report estimates of overall global energy consumption from data centers based on what they consider to be high quality studies based on three methodologies: global estimates, country/regional estimates, and estimates based on large companies. All studies they consider report worldwide energy usage between 210 and 470 terawatt-hours.

The above numbers comprise just over 1% of world electricity consumption. Some studies report much higher data center energy consumption, but they are judged to be of poor quality 3. Projections for 2030 range from 200 to 8000 TWh per year, though the larger estimates are from studies judged to be of lower quality 3.

In the 2020s, the biggest driver of data server energy growth has been artificial intelligence and generative AI in particular. Most studies considered found that energy usage associated with AI was less than 75 TWh per year in 2023 3.

Researchers at Lawrence Berkeley National Laboratory 4 have estimated the following United States data center energy usage for 2023.

Source: Shehabi et al. 4.

Energy Efficiency

Masanet et al. 5 consider energy consumption in data centers from 2010 to 2018. Over that time, electricity consumption increased from 194 TWh to 205 TWh, a 6% increase, while compute instances increased by a factor of 6.5. Some higher forecasts for future data center electricity demand fail to account for future efficiency gains 3.

The energy efficiency of data transmission have been observed to roughly double every two years per gigabyte 6.

Energy Savings Potential

There are several options available to reduce the energy needs of data centers. European data centers have an average Power Usage Effectiveness--the ratio between total energy consumption and that used directly by the processors--of 1.8 7. The state of the art is at worst 1.12, indicating a potential savings of 38%. Data centers are typically more energy efficient when they can use the air of a cooler climate for cooling 8. Eventually, liquid cooling of processors could save 80% of the data center's energy consumption.

However, the rebound effect--the tendency for energy efficiency gains to be partially or entirely offset by increased consumption--is especially applicable to computation and data transfer 9. Energy efficiency is a key enabler of emerging information technologies such as deep learning, virtual and augmented reality, blockchain, and autonomous vehicles.

Video Streaming

The energy requirements of video streaming are estimated as follows.

The image: "streaming_emissions.svg" cannot be found!

Streaming emissions, as estimated by George Kamiya (10 via 11). Emissions are direct emissions, not accounting for embodied energy within devices or equipment. For streaming and the kettle scenario, the world average electricity mix is assumed, and for driving, the world average internal combustion car is assumed. For the smartphone, tablet, laptop, and TV scenarios, streaming is assumed to be done over a high-powered WiFi connection, whereas for the average scenario, steaming occurs over the average network connection. Because 4G connections are typically less energy-intensive than WiFi, the average data trasmission energy requirement is less than for any individual device. Figures are averages as of 2019, and due to rapid changes in the IT industry, figures may now differ significantly 12.

Significantly higher steaming energy figures, often reported in the press, generally derive from an analysis done by The Shift Project in 2019 13. We believe Kamiya's figures are more reliable, as The Shift Project's figures for energy and carbon intensity of streaming are outdated and unrealistically high.

The growing popularity of virtual reality gaming 14; the commercialization of 8K 15 and higher resolution 16 displays, which contain four or more times as many pixels as UHDTV displays noted above; cloud gaming 17; the rollout of 5G networks 18; and the recent surge in demand for video conferencing 19 are among the factors that are likely to drive high demand growth for video streaming in the near future.

Cryptocurrency

A cryptocurrency is a digital asset for which ownership and transactions are represented in a ledger using strong cryptography, typically a blockchain. Of many cryptocurrencies now in circulation, Bitcoin is the first and by far the largest by market capitalization 20. A main goal of Bitcoin is to develop a system of transaction that is free of centralized control from governments and corporations 21. Bitcoin has come under criticism for the heavy energy consumption of its proof-of-work mining system.

The image: "bitcoin_energy.svg" cannot be found!

Estimated annualized energy consumption for Bitcoin and Ethereum was accessed on March 9, 2021. At this time, the prices of Bitcoin (about $54,000) and Ethereum (about $1800) were near record highs, and thus energy consumption is higher than normal. It should be noted that, while assessed aspects of traditional banks have energy consumption comparable to Bitcoin, banks serve far more customers worldwide. Sources: 22, 23, 24, 25, 26. By way of comparison, the world uses about 27,000 TWh of electricity per year.

As of 2020, 39% of proof-of-work mining was powered by renewable energy 27, as miners take advantage of low prices resulting from excess capacity.

It is possible to save energy within Bitcoin and other cryptocurrencies by moving some transactions off the main chain, such as through the Lightning Network 28. Proof-of-stake blockchains typically require less energy than proof-of-work blockchains 29.

Machine Learning

Machine learning, a subfield of artificial intelligence, is a technique to develop algorithms through data. Deep learning, a type of machine learning that is based in neural networks, has in particular advanced considerably over the past decade 30. Advances in machine learning have underpinned areas such as natural language processing; gesture, audio, and video recognition; and robot locomotion, and its importance could grow greatly in the coming years. However, the energy needs of machine learning today and possible energy needs in the future are growing concerns.

The image: "ml_energy.svg" cannot be found!

Source: Biewald 31. By way of comparison, the world uses about 27,000 TWh of electricity per year.

Training a state-of-the-art deep learning model is expensive in terms of energy and CO₂ emissions, and the cost makes it prohibitive for any but large institutions to do cutting-edge machine learning research.

The image: "ml_energy2.svg" cannot be found!

Most figures are from Strubell et al. 32, with ELMo and BERT described by Peters et al 33 and Devlin et al. 34 respectively. Strubell et al. examine GPT-2, OpenAI's state-of-the-art NLP model at the time, and some other models based on tensor processing units, but energy and emissions figures are not available.

From 2012-18, the compute in training runs of AI models doubled every 3.4 months 35, much faster than the 2 year doubling time of transister density that defined Moore's Law (a trend that itself may be faltering), with the expectation that models will continue to grow rapidly.

The image: "model_size.svg" cannot be found!

The number of parameters of a model is a reasonably good proxy for the computer required to train it. Sources: most figures reported by the Allen Institute 36, with ELMo developed by Peters et al. 37 and GPT-3 by OpenAI 38.

The rapidly growing size of deep learning models, together with diminishing returns with model size, have inspired a Green AI methodology, which would make energy and cost efficiency a metric by which models are evaluated, along with accuracy on a test set 36.

References

  1. Spencer, T., Singh, S., D'Ambrosio, D., Hopewell, H., Jacamon, V., Martinos, A., Salmon, N., Wanner, B. et al. "Energy and AI". International Energy Agency. April 2025. ↩ ↩2

  2. Energy Institute. "Statistical Review of World Energy". 2024. ↩

  3. Kamiya, G., Coroamă, V. C. "Data Centre Energy Use: Critical Review of Models and Results". Efficient, Demand Flexible Appliances, Energy Efficient End-use Equipment, International Energy Agency Technology Collaboration Programme. March 2025. ↩ ↩2 ↩3 ↩4 ↩5

  4. Shehabi, A., Hubbard, A., Newkirk, A., Lei, N., Siddik, M.A., Holecek, B., Koomey, J., Masanet, E., Sartor, D. "2024 United States Data Center Energy Usage Report". Lawrence Berkeley National Laboratory, Energy Analysis & Environmental Impacts Division. 2024. ↩ ↩2

  5. Masanet, E., Shehabi, A., Lei, N., Smith, S., Koomey, J. "Recalibrating global data center energy-use estimates". Science 367(6481), pp. 984-986. February 2020. ↩

  6. Aslan, J., Mayers, K., Koomey, J. G., France, C. "Electricity Intensity of Internet Data Transmission: Untangling the Estimates". Journal of Industrial Ecology 22(4), pp. 785-798. August 2017. ↩

  7. Avgerinou, M., Bertoldi, P., Castellazzi, L. "Trends in Data Centre Energy Consumption under the European Code of Conduct for Data Centre Energy Efficiency". Energies 10(10), 1470. September 2017. ↩

  8. Song, Z., Zhang, X., Eriksson, C. "Data Center Energy and Cost Saving Evaluation". Energy Procedia 75, pp. 1255-1260. August 2015. ↩

  9. Mills, M. "Energy and the Information Infrastructure". Real Clear Energy. September 2018. ↩

  10. Kamiya, G. "Factcheck: What is the carbon footprint of streaming video on Netflix?". Carbon Brief. February 2020. ↩

  11. Kamiya, G. "The carbon footprint of streaming video: fact-checking the headlines". International Energy Agency. December 2020. ↩

  12. Fulton, S. "How Much Is Netflix Really Contributing to Climate Change?". Data Center Knowledge. February 2020. ↩

  13. Efoui-Hess, M. "The Climate Crisis: The Sustainable Use of Online Video". The Shift Project. July 2019. ↩

  14. KÖLÜŞ, C., BAŞÇİFTÇİ, F. "The Future Effects of Virtual Reality Games". Recent Research and Assessments for Computer Engineering, Academic Works of Livre de Lyon. 2020. ↩

  15. Mordor Intelligence. "8K Market: Growth, Trends, COVID-19 Impact, and Forecasts (2021-2026)". ↩

  16. Alvarez-Mesa, M., Sanz-Rodríguez, S., Chi, C. C., Glowiak, M., Haring, R. "8K/16K Video and 3D Audio Coding and Playback for Large-Screen Immersive Spaces". SMPTE Motion Imaging Journal 130(1), pp. 50-58. January 2021. ↩

  17. Markets and Markets. "Cloud Gaming Market by Offering (Infrastructure, Gaming Platform Services), Device Type (Smartphones, Tablets, Gaming Consoles, PCs & Laptops, Smart TVs, HMDs), Solution (Video Streaming, File Streaming), Gamer Type, Region - Global Forecast to 2024". 2019. ↩

  18. Jha, S. "Impact of 5G on OTT and Broadcasting". Simnovus. Accessed March 10, 2021. ↩

  19. Wang, S. S., Roubidoux, M. A. "Coronavirus Disease 2019 (COVID-19), Videoconferencing, and Gender". Journal of the American College of Radiology 17(7). July 2020. ↩

  20. Nakamoto, S. "Bitcoin: A Peer-to-Peer Electronic Cash System". bitcoin.org. October 2008. ↩

  21. Dodd, N. "The social life of Bitcoin". Theory, Culture & Society 35(5), pp. 35-56. December 2017. ↩

  22. Cambridge Center for Alternative Finance. "Cambridge Bitcoin Electricity Consumption Network". Accessed March 9, 2021. ↩

  23. Digiconomist. "Bitcoin Energy Consumption Index". Accessed March 9, 2021. ↩

  24. Digiconomist. "Ethereum Energy Consumption Index (beta)". Accessed March 9, 2021. ↩

  25. Domingo, C. "The Bitcoin vs Visa Electricity Consumption Fallacy". Hackernoon. November 2017. ↩

  26. Zodhya. "Which consumes more power: Banks or Bitcoins?". medium.com. April 2018. ↩

  27. Blandin, A., Pieters, G., Wu, Y., Eisermann, T., Dek, A., Taylor, S., Njoki, D. "3rd Global Cryptoasset Benchmarking Study". University of Cambridge Judge Business School. SSRN 3700822. September 2020. ↩

  28. Robert, J., Kubler, S., Ghatpande, S. "Enhanced Lightning Network (off-chain)-based micropayment in IoT ecosystems". Future Generation Computer Systems 112, pp. 283-296. November 2020. ↩

  29. Zhang, R., Chan, W. K. "Evaluation of Energy Consumption in Block-Chains with Proof of Work and Proof of Stake". Journal of Physics: Conference Series 1584: 012023. 2020. ↩

  30. Krizhevsky, A., Sutskever, I., Hinton, G. "ImageNet Classification with Deep Convolutional Neural Networks". Advances in neural information processing systems 25, pp. 1097-1105. 2012. ↩

  31. Biewald, L. "Deep Learning and Carbon Emissions". Toward Data Science. June 2019. ↩

  32. Strubell, E., Ganesh, A., McCallum, A. "Energy and Policy Considerations for Deep Learning in NLP". Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics: System Demonstrations. July 2019. ↩

  33. Peters, M. E., Neumann, M., Iyyer, M., Gardner, M. "Deep contextualized word representations". ArXiv, Allen Institute for Artificial Intelligence. March 2018. ↩

  34. Devlin, J., Chang, M., Lee, K., Toutanova, K. "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding". ArXiv, Google AI Language. May 2019. ↩

  35. OpenAI. "AI and Compute". May 2018. ↩

  36. Schwartz, R., Dodge, J., Smith, N. A., Etzioni, O. "Green AI". ArXiv, Allen Institute for Artificial Intelligence. August 2019. ↩ ↩2

  37. Peters, M. E., Neumann, M., Iyyer, M., Gardner, M., Clark, C., Lee, K., Zettlemoyer, L. "ELMo". Allen Institute for Artificial Intelligence, NAACL 2018. 2018. ↩

  38. Brown, T. et al. "Language Models are Few-Shot Learners". ArXiv, OpenAI. ↩