Does AI Carry a Hefty Economical and Environmental Price Tag?

AI is all the rave, and to be honest, our team at Pinnacle IT finds it fascinating. Aside from the temptation to fall pray to the sci-fi fears of AI taking over the world, (sorry, I just don't see it), it's been difficult to see any real downside to the new push for the development and use of artificial intelligence.

Until now. We all know there has to be a price. The tools at our disposal made available by the power of AI cannot come at zero cost; it's just too good to be true.

So here it is. One of the impacts of artificial intelligence is the crazy amount of power required to uphold the infrastructure for these systems. OpenAI's GPT-4 is powered by about 25,000 Nvidia A100 GPUs. This requires around 60,000 megawatt hours. Basically, you could power a small city's households for 5 years or more.

And that's only the one project. Other AI superstars employ the use of 350,000 Nvidia H100 GPUs!

This is a massive resource burden, and it's going to take a whole lot of dollars to keep it up. Can you imagine the environmental impact of all of this?

Visual Capitalist provides this overview of Microsoft’s rising electricity needs as it works with OpenAI on its own project.