Electricity demand to double in 3 years. How AI and mining play a part

Electricity demand worldwide could double over the next three years, mainly due to cryptocurrency mining and artificial intelligence. How will crypto keep up?
AI and cryptocurrencies accounted for almost 2% of global electricity demand in 2022, which illustrates the scale of their energy impact. This increase is mainly due to the growing complexity and volume of computing operations for artificial intelligence and the ever-increasing number of cryptographic transactions.
According to a recent report from Bloomberg, which cites the International Energy Agency, global demand for electricity from data centers, cryptocurrencies, and artificial intelligence could more than double over the next three years, amounting to the equivalent of Germany’s entire electricity demand.
Source: International Energy Agency

AI’s appetite to grow

Artificial intelligence (AI) has become an integral part of modern life, streamlining various aspects. However, its extensive integration raises concerns about a substantial surge in energy consumption. According to Alex de Vries, a graduate student at the University of Amsterdam Free University, the global AI infrastructure might demand an energy equivalent to that of an entire country.
It’s been nearly a year since the AI industry entered a phase of rapid expansion triggered by the introduction of the OpenAI ChatGPTAI chatbot. The training and operation of neural networks, such as the one powering this service, involve a notably energy-intensive process. Hugging Face, an AI-developing company and a key contributor to large language models, disclosed that training its platform demanded 433 megawatt-hours (MWh) of electricity – roughly equivalent to the energy needs of 40 average American households for a year. In comparison, ChatGPT, being a more extensive project, consumes approximately 564 MWh daily.
And there seems to be no way out of this situation. Companies worldwide are working to improve AI systems’ efficiency in hardware and software, making them less energy-intensive. However, increasing the efficiency of these machines will inevitably increase the demand for AI services, causing resource use to increase even further—a phenomenon in economics known as Jevons’ Paradox.
You might also like:
2050: CBDCs, AI, and the uncharted path ahead | Opinion 

How AI will change electricity demand

Amid these developments, the most substantial projects raise concerns. According to the paper by Alex de Vries, Google currently handles a staggering 9 billion search queries daily. Using this information, the researcher calculated that if AI were employed for every Google search, it would demand approximately 29.2 terawatt-hours (TWh) of energy annually – a figure on par with Ireland’s total consumption. While de Vries acknowledges that this extreme scenario may not materialize in the near future, he does concede that the expansion of AI servers will contribute to an overall rise in energy consumption. By 2027, the cumulative volume could range between 85 and 134 TWh annually.
This volume is comparable to the needs of larger countries, such as Argentina, the Netherlands, and Sweden. Increasing the efficiency of AI accelerators will allow developers to repurpose processors to solve problems associated with AI algorithms, which will additionally lead to an increase in the related energy consumption.
The future of artificial intelligence, where everyone has a personal assistant, relies heavily on having access to affordable energy. Without it, the pace of AI progress could slow down considerably. There’s hope for cost-effective energy sources through innovations like new energy solutions or advancements in thermonuclear fusion. Currently, creating robust neural networks, exemplified by projects like ChatGPT, is a privilege limited to a handful of major players with hefty investments, such as OpenAI.