As the International Energy Agency (IEA) forecasts, AI’s energy consumption is poised for an explosive increase, overshadowing even the substantial growth in crypto’s energy use. Despite AI’s burgeoning demand, the spotlight remains on cryptocurrency for its significant energy footprint.
The global electricity market is going through major changes.
This is mainly because the energy industry is at the forefront of reducing carbon emissions and adapting to new ways of using power.
Two areas that are expected to change how much energy we use are cryptocurrency mining and artificial intelligence (AI). The International Energy Agency (IEA), a big organization made up of many countries, believes these changes will happen in the next few years.
According to the IEA’s report for 2024, which offers a forecast until 2026, there’s some good news about energy use.
Right now, making electricity is the biggest reason for carbon dioxide (CO2) emissions worldwide. But, this sector is also leading the charge towards zero emissions.
By 2025, renewable sources like wind and solar are expected to be the main way we generate power.
Energy use increased at a slower rate, going from 2.4% in 2022 to 2.2% in 2023. However, it’s predicted to jump to 3.4% by 2026. This increase will mainly come from countries like China and India.
At the same time, the amount of energy used by data centers, artificial intelligence, and cryptocurrency is expected to more than double during this period, going over 1,000 terawatt-hours (TWh). Notably, a third of all types of data centers are found in the United States.
According to the IAE rapport, electricity consumption is expected to rise up to 1,000 TWh by 2026.
Source: IAE
Artificial Intelligence (AI) is set to lead in energy consumption, with its usage predicted to increase by ten times from 2023 to 2026. For example, just ChatGPT is expected to use almost 10 terawatt-hours (TWh) each year during this period.
To put that in perspective, every time someone uses ChatGPT, it uses about ten times more energy than a single Google search.
Bitcoin energy consumption was measured to be about 120 TWh in 2023.
This was part of the 130 TWh used for all cryptocurrency mining that year, which increased from 110 TWh in 2022.
At that time, cryptocurrency mining made up 0.4% of the world’s total energy consumption. The International Energy Agency (IEA) forecasts that by 2026, cryptocurrency mining will consume 160 TWh.
Even though the situation with energy use is changing and cryptocurrency mining only uses a small part of the world’s energy, the report points out that crypto is still worrying:
The report notes that it’s hard to reduce electricity use because any energy saved might just be used up by other things that need a lot of power, like different types of cryptocurrencies besides Bitcoin, even though some are getting more efficient.
It’s said that Bitcoin mining uses 54.5% sustainable energy. There’s an increase in mining activity as the Bitcoin halving approaches, leading many miners to invest a lot in new equipment.
Source: IEA
To put things into perspective, let’s see what 1 TWH means in terms of normal consumption.
One terawatt-hour has enough energy to supply electricity to 70,000 homes in the United States for a whole year.
When comparing this to the energy use of entire countries, it means that the combined energy consumption of cryptocurrency and related users is expected to reach the same level as Japan’s consumption.
How can energy consumption be more efficient?
To make energy consumption more efficient, especially in data centers, several strategies and regulations are crucial.
Regulatory and policy initiatives
- Regulatory measures. The European Commission’s revised Energy Efficiency Directive imposes regulations on the European data center sector to enhance electricity demand management. Starting in 2024, data center operators are required to report their energy use and emissions. Large-scale data centers must also incorporate waste heat recovery applications, when feasible, and aim for climate neutrality by 2030.
- Efficiency standards. An EU regulation effective since 2020 sets efficiency standards for data centers, helping to control their environmental impact. The Climate Neutral Data Centre Pact, a self-regulatory initiative, aims for climate neutrality in the sector by 2030.
- U.S. energy policies. The Energy Act of 2020 in the United States mandates studies on energy and water use in data centers. It encourages the development of efficiency metrics and best practices. The Department of Energy (DOE) is also focusing on producing more efficient semiconductors, which can reduce cooling requirements.
- Chinese regulations. China requires data centers acquired by public organizations to improve energy efficiency and be fully powered by renewable energy by 2032, starting with a renewable energy share mandate of 5% in 2023.
Technological and operational innovations
- Advanced cooling systems. Adopting high-efficiency cooling systems can significantly reduce electricity demand. Innovations like direct-to-chip water cooling and the use of specific low viscous fluids are promising.
- Machine learning and AI. Utilizing machine learning, as Google did with its DeepMind AI, can optimize data center operations, significantly reducing electricity demand.
- Quantum computing. In the long term, replacing supercomputers with quantum computers, which require less energy, could reduce overall electricity demand. However, efficient cooling systems are necessary due to their low operating temperatures.
- Hyperscale data centers. Transitioning to Hyperscale Data Centers, which can handle large-scale operations without significantly increasing electricity consumption, is both sustainable and financially attractive.
- Carbon-aware models. Software that allows for time and location shifting of electricity demand to regions with lower carbon intensity can significantly reduce emissions and operational costs.
By combining these approaches, data centers can significantly increase their share of carbon-free energy in total electricity consumption, as demonstrated in Google’s 2023 Environmental Report.