Data centres and AI. A combination often presented as the infallible recipe for an increasingly prosperous and decarbonised future. Reality is different, the future looks bleak, surely not brighter.
Every prompt, every request to AI, is a new and huge demand for energy. Far more than imagined. Perhaps it would be a good idea to equip every smartphone or PC with a meter or an LED that displays the energy cost and carbon footprint associated with each request.
But this won’t happen. It would stimulate sensible thinking, which also means curbing usage, but profit comes from expanding usage. Nearly all of the AI computational power is used to produce real-time responses, thereby expanding the consumer base, which is the ultimate goal of big tech behind AI.
Presented almost as a game and perceived as a useful shortcut, artificial intelligence – thanks in part to its forced integration into search engines and smartphones – has had no trouble establishing itself in everyday practices, both for individuals and within businesses. This rapid, ever-accelerating spread has brought with it the need for ever-larger data centres and, therefore, a disproportionate amount of additional energy, water, and space. Resources that are required to house and maintain a technology still in its infancy, with enormous scope for improvement.
Despite these issues, data centres are presented as enablers of development and progress. At least, by those who tout them as the salvation for defunct activities like coal mines that are no longer useful or viable for extraction, but which can provide the ideal conditions for housing these machines, which require space and low temperatures. The only bright spot is for the builders and the construction market as a whole.
How much energy and natural resources are actually used to write a text, create an image or generate a video?
It is by no means straightforward to determine what is being requested. A video, a translation, a solution, an answer to a simple geographical or historical question, a greeting—how many parameters are in the initial prompt? At what point is the request submitted? Which data centre processes the request? What energy source powers the data centre? Fossil fuels or renewables? The answer is mainly fossil fuels, because data centres cannot afford intermittent energy—and this is driving the hi-tech sector’s push for a revival of nuclear power.
The honeymoon period for data centres is over. Every day, there is a new local community that opposes a data centre project, being only too aware that it drives up local electricity demand, pushes up power prices for households and businesses, increases pressure on power and water systems, and raises CO2 emissions.
Nothing can match the internet and social media in fueling opposition to data centres and circulating news about the impact these facilities have on the environment and local communities where they are located. Opposition is growing. The irony is that someone still asks AI how to fix AI.
Gianni Serra
-
By: ONE Editor
Italian professional journalist. ONE Editor in Chief and Sotacarbo Director of Communications and International Relations.
-
