AI – and with it, HR tech – are going for the nuclear option.
No, I don’t mean the whole industry is about to go belly up (or at least I hope not), but in the literal sense of the word. Late last year, major AI and technology companies announced their bold plans to invest in nuclear technology to meet the growing energy demands of their industry.
There has already been an avalanche of negative press about the energy demands of AI, or even the energy demands of a single query on an AI platform. The infrastructure that powers AI, such as servers and data centers, is placing a significant strain on the energy grids of many states, leading to increased market-rate controlled energy prices and, in turn, higher residential energy bills.
This has left many furious and is another blow to public sentiment toward Big Tech and AI at a time when many already feel that Big Tech is hoarding wealth, surveilling them, and potentially bringing in the Anti-Christ (see Peter Thiel). With energy costs now being passed on to residents, the optics for those many citizens footing the bill are that they’re being forced to subsidize the wealthiest companies on the planet, which they view as using that energy primarily to automate or replace their jobs.
Setting the Stage
Let’s first understand the problem: AI requires a huge amount of computation, which demands much more electricity than previous web-based technologies. Until recently, computational electricity demand grew gradually. From 2000 to 2010, US consumption rose from ~30 TWh to ~70 TWh, adding roughly the population of Belarus (9 million people) to the grid over 10 years. Between 2010 and 2017, data center electricity use remained stable.
That all changed with the rise of generative AI. A single ChatGPT reply can involve trillions of calculations, making it nearly 10 times more energy-intensive than a Google search, and generating images demands even more energy. From 2018 to 2023, data center energy use rose from 1.9% of the total US electricity use to 4.4%. This growth is comparable to adding the entire energy needs of Bangladesh (~175 million people) to the US power grid within only five years.
Continue Reading: