AI's Secret Power Thirst: Agentic AI is Building a New Energy Crisis

You've probably noticed the buzz about massive data center buildouts, and if you're wondering why, given AI's current capabilities, you're onto something big. The answer isn't about powering your average recipe search or travel query. Those simple chatbot interactions are becoming an outdated way of looking at AI.

We're on the cusp of, or rather, deep into, the age of agentic AI , and it's a game-changer for energy demands.

What Exactly Are AI Agents?

Think beyond asking a chatbot a simple question and getting an answer.

Agentic AI refers to large language model-based systems designed to make autonomous decisions and execute complex tasks.

Instead of just replying, an agent can break down a goal, generate hundreds of sub-prompts, and work tirelessly behind the scenes to achieve it.

My colleague Maxwell Zeff puts it perfectly: 'For example, if someone asked an AI agent to build them a website, it might run for hours to build out features, re-prompting itself dozens of times in the process to build different web pages, menus, and datasets that power the thing.' This isn't a quick search; it's a dedicated, resource-intensive project.

The Astronomical Energy Bill

Frontier AI labs are heavily invested in agents. Recently, OpenAI revealed a swarm of over 10,000 agents sent 2.7 million messages to solve a complex math problem. While mathematicians debated the claims, the sheer scale of computation involved is undeniable.

Max tells me this likely cost tens of millions of dollars in processing power - a staggering amount for a single endeavor.

Historically, private AI companies have been less than transparent about their environmental footprint. Many CEOs try to minimize concerns by pointing to the energy use of single queries. OpenAI CEO Sam Altman, for instance, once controversially equated the water needed for a single almond to 38,000 ChatGPT queries.

But agentic AI blows these comparisons out of the water.

The tasks can range from simple automation to full days of autonomous coding, often involving teams of parallel 'helper' agents.

Decoupled from Humanity?

This shift creates a massive gulf in power consumption. Unlike other tech growth areas, where usage is tied to human actions, like people streaming Netflix or driving cars, agentic AI can operate independently. Boris Gamazaychikov, co-founder of Sustainable AI Group, notes, 'Now, this stuff is kind of decoupled from users, and if you listen to AI leaders, I think that's what they want.

They're talking about unicorns that have one employee .' That 'one employee' would be human, while hundreds or thousands of AI agents do the heavy lifting.

This vision of AI, where agents work autonomously at scale, means an almost limitless expansion of energy demand as tasks grow more complex. The question isn't if these data centers are needed, but what the true cost of this AI-driven future will be for our planet and our power grids.