Is AI Too Thirsty for Its Own Good?

See All Articles

The Unseen Problem with AI: It Needs How Much Power?!
We're all pretty amazed by Artificial Intelligence these days, aren't we? From writing emails and helping doctors to even composing music, AI seems to be everywhere, constantly getting smarter. But what if the biggest hurdle for AI's future isn't about making it smarter, but simply keeping it on?
That's the stark warning coming from Eric Schmidt, the former CEO of Google. He recently dropped a bombshell: the real limit to how much AI can grow isn't about building better computer chips, but about having enough electricity to power them.
AI's Insatiable Appetite
Think about it: all those incredible AI programs run on massive data centers, which are essentially giant warehouses full of super-powerful computers. As companies like Google, Microsoft, and OpenAI race to create "Artificial General Intelligence" (AGI) – AI that can think and reason like humans, or even better – these data centers are sucking up energy at an alarming rate.
Schmidt, who now leads a pro-AI think tank, explained just how big this problem is. He estimates that the U.S. alone might need an extra 92 gigawatts of power to keep up with its AI ambitions. To put that in perspective, that's like building 92 brand new nuclear power plants! And for a bit more perspective, the U.S. has only managed to build two such plants in the last three decades. We need a lot of power, and we need it fast.
More Than Just Electricity: The Water Problem
It's not just about plugging in. These powerful AI computers generate a lot of heat, so they need massive amounts of cooling. And that means they need water. Lots of it.
Companies are already scrambling. Microsoft, for example, has signed a huge 20-year deal to get power from a nuclear plant that was previously shut down. Sam Altman, the head of OpenAI, is even investing in companies trying to create fusion energy – a futuristic, super-clean power source.
The numbers are eye-opening: Microsoft's own report showed a 34% jump in water use in 2023, guzzling 1.7 billion gallons just to cool its AI data centers. Researchers even predict that by 2027, AI could demand enough water to supply all of Canada for an entire year!
The Big Dilemma
This massive energy and water demand is ringing alarm bells for environmental groups like Greenpeace, who worry it could derail our efforts to fight climate change.
It's a strange twist: AI promises to solve some of the world's biggest problems, from medical breakthroughs to climate solutions. Yet, its very development could strain – and possibly break – the energy and water systems it aims to improve.
As Schmidt puts it, "We don’t know what superintelligence will ultimately deliver, but we know it’s coming. And we must plan now to make sure we have the energy infrastructure to support it."
So, while we marvel at how smart our machines are becoming, perhaps the more urgent question is: will we have enough power to keep them running?

Read more