Are there some important “details” about artificial intelligence (AI) and cryptocurrencies that most uninitiated people don’t know?
Obviously many – among them the fact that these are “giants” who insatiably devour huge amounts of energy, but also water to cool their systems. They do so mainly through databases, which are already an integral part of the backbone of the digital age humanity has entered, while it is certain that their role will be further upgraded.
What exactly this means and what we should expect in the coming years, against the background of the leaping development of artificial intelligence and its tools, but also the position that cryptocurrencies have in the markets, he explains to us comprehensively and without leaving much room for doubt the International Energy Agency (IEA), in its recent report: “Global electricity consumption by databases, AI and the crypto-currency industry is likely to double by 2026,” it says characteristically, leaving no room for misinterpretation.
The consequences
But what will be, in practice, the consequences of all the above? Quite simply, as the IEA experts explain, that of the 460 terawatt hours (TWh) of power estimated to have been consumed during 2022 (for which there are comprehensive data), the consumption in 2026 is estimated to reach even 1,000 terawatt hours .
This is a quantity that will be equivalent to the total electricity needs of the second largest economy on the planet, Japan, while only the expected increase will be as if one had added to the grid a country like Sweden (in the mild scenario) or Germany (at worst). It will represent 5% of global demand, compared to “only” 2% in 2022.
The Irish model
It is worth noting that most of the existing bases are located in the cities that are considered financial centers – Frankfurt, Paris, Amsterdam and Dublin. Especially for Ireland, the picture is more than impressive: Based on the IEA’s estimates, from 17% of the electricity consumption absorbed in 2022 by databases in the country to operate, within the next three years the percentage will double , reaching 32%.
Because many may still doubt the share of “responsibility” that artificial intelligence has in this development, the IEA also has the answer: “The rapid integration of AI into software programming in a number of fields is increasing the demand for electricity in the databases” she notes in her report.
In fact, it gets even more specific, giving the following example: Search engines, such as Google’s, are likely to see their energy consumption increase tenfold if they decide to fully integrate AI. And that’s because comparing the average consumption in a standard search (about 0.3 watt hours, Wh), with the corresponding one in a search through OpenAI’s ChatGPT (almost 2.9 Wh) and considering at the same time that there are about 9 billion searches per day , the result translates into an additional 10 terawatt hours of consumption each year.
New techniques
Something similar happens with cryptocurrencies, such as Bitcoin and Ethereum. In 2022, in particular, they are estimated to have consumed 110 TWh, which alone accounted for 0.4% of total global demand – thus absorbing roughly as much electricity as the Netherlands needed in the same year. For 2026, based on the “base model” of the relevant chart, cryptocurrencies will absorb 160 TWh – even if the most optimistic believe that by applying new techniques for their “mining” the amount can be significantly reduced (already Ethereum managed to reduce it by 99% in 2022).
The question that arises and must be answered by all is, therefore, a reasonable one: Is the estimated benefit worth the cost that will be incurred?
Problem with AI
In his speech at this year’s World Economic Forum in Davos, the CEO of OpenAI and a Microsoft executive Sam Altman effectively admitted that the AI industry is headed for an energy crisis. In particular, it warned that the new generation of its tools will consume unimaginably larger amounts of energy than originally calculated, which means that if effective solutions are not found, the resulting problem will be big.
Clearly, the problem will not only concern the AI industry, as the entire planet faces an uncertain energy future. It is not paradoxical, then, that intelligence predicts that artificial intelligence, and especially the energy side it needs, were central issues at the UN’s international conference on climate change, COP28, held last year in the UAE, while something similar is expected to happen with the next one, COP29 in Baku.
4% of world consumption
According to existing data, in 2022 there were around 8,000 databases worldwide, with 33% of them in the US, 16% in Europe and 10% in China. Especially with regard to the member countries of the European Union, the consumption of electricity derived from the databases amounted in that particular year to 100 terawatt hours (TWh), a quantity that represented approximately 4% of the total, while until 2026 and based on planning that exists, it is estimated that it will rise to 150 TWh, now representing a percentage of more than 5%.
Coordination: Angelos Skordas
Diligence: Panagiotis Sotiris
#hungry #energy #Artificial #Intelligence #cryptocurrencies
