There was a time when the electricity for running server farms was an afterthought, but in many place, that time is already past. The International Energy Agency has a discussion of the issue in its recent report, Electricity
2024: Analysis and forecast to 2026 (January 2024, pp. 31-36).
The light blue and dark blue bars show the rise in electricity from 2022 to 2026 needed for server farms for the US, EU, and China, together with Denmark and Ireland. The expected rise is substantial. But focus instead on the orange diamonds, which show the share of total electricity for the country or region as a whole. For the US and the EU, server farms are projected to be 5-6% of total electricity consumption by 2026. For Denmark, the projection for 2026 is about is 20% of total electricity consumption; for Ireland, it’s 32%.
Ireland’s situation is unique: it has both a welcoming business climate for foreign investment and a geographically advantageous position between the US and EU economies. But its experience also illustrates that data centers tend to be concentrated in a few locations, often locations of tech firms, financial centers, and government. For the US, the IEA report notes:
In the United States, the largest data centre hubs are located in California, Texas and Virginia. In the case of Virginia, their economy was dominated in 2021 by the data centre sector expansion, attracting 62% of all of the state’s new investments and providing more than 5 000 new jobs. Northern Virginia is the largest data centre market in the country, collecting USD 1 billion in local tax revenues per year, with growth trending higher as companies, such as Amazon’s planned USD 35 billion expansion by 2040, continue to increase their investment in the state.
Of course, there are potential innovations with some promise for holding down the increase in this level of electricity use–but the fundamental drivers behind the rising demand for data servers seem likely to keep rising. Here’s one vivid example from the IEA report:
Market trends, including the fast incorporation of AI into software programming across a variety of sectors, increase the overall electricity demand of data centres. Search tools like Google could see a tenfold increase of their electricity demand in the case of fully implementing AI in it. When comparing the average electricity demand of a typical Google search (0.3 Wh of electricity) to OpenAI’s ChatGPT (2.9 Wh per request), and considering 9 billion searches daily, this would require almost 10 TWh of additional electricity in a year.
Providing the additional electricity for data centers will be a challenge, but my guess is that, when it comes to the everyday problems of business and government as well as the big-picture problems of national and global policy-making, additional data-processing capacity is a (relatively) cheap investment.
