The Power-Hungry Future of AI Agents

The Rise of Agentic AI

What exactly are we building all these data centers for? That's a question that's been on many minds lately, especially as tech giants pour billions into constructing some of the world's most power-hungry facilities. It's not just about simple chatbot queries anymore—AI has evolved dramatically, moving toward what we now call agentic AI systems.

These aren't your average AI assistants responding to a single prompt. Agents are complex systems designed to autonomously carry out tasks by generating multiple sub-tasks and even collaborating with other agents in real time. In short, they operate with increasing independence, often performing actions that would take humans hours or days—without any direct human intervention.

"Rather than asking an AI chatbot a simple question and answer, these agents can give themselves hundreds of small prompts based on a user's original question," explains my colleague Maxwell Zeff. "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."

This transformation marks a pivotal shift in how we think about artificial intelligence and its impact on infrastructure.

Energy Demands: A Hidden Cost

But this new paradigm comes with a steep price tag—literally. The computational power required to run even a modest number of AI agents is enormous. For instance, OpenAI recently reported that over 10,000 agents generated 2.7 million messages in pursuit of solving a complex math problem.

While such breakthroughs are exciting, they also consume staggering amounts of electricity. It's hard to quantify exactly how much energy this represents, but estimates suggest tens of millions of dollars' worth. And while AI labs may not be forthcoming with details, it's clear that these agents are far more energy-intensive than simple user queries.

This is where the conversation becomes particularly important for those of us who track economic trends and their human impact. When we consider the massive investments being made in data centers, it's not just about profits—it's about ensuring sustainable growth and responsible resource use.

Climate Implications

As we continue to expand our digital footprint, so too does the environmental toll. Climate scientist Zeke Hausfather recently calculated that his average daily Claude session consumes as much energy as two refrigerators running continuously.

While this may sound like a small figure compared to other household activities, it adds up when multiplied across millions of users. And as platforms like Meta begin rolling out personal agents—like Muse, designed for use with AI glasses and cloud-based processing—it becomes increasingly likely that everyday individuals will be unknowingly contributing to a growing carbon footprint.

"The technology that's going to be trained by the data centers that are being proposed and built right now is three to five years away," says Boris Gamazaychikov, CEO of Sustainable AI. "It's going to be a very different flavor than just the chatbot window."

This future could mean an explosion in energy demand, especially if AI agents become embedded in daily life through wearables and smart devices.

Infrastructure Solutions and Challenges

In response to this growing power need, tech companies are racing to secure reliable sources of electricity. Some are turning to small modular reactors (SMRs) as a potential solution—small nuclear plants that could offer clean, consistent energy for large data centers.

However, despite decades of development, SMRs have yet to see widespread commercial adoption in the U.S. The Trump administration's efforts to fast-track licensing may help accelerate progress, but many developers are unwilling to wait years for such technologies to mature. Instead, they're relying on natural gas turbines—more immediate but less environmentally friendly solutions.

The result? A patchwork of data center construction projects that reflect both urgency and uncertainty in our approach to energy management.

Policy Shifts and Public Reaction

Not everyone is pleased with the scale of this expansion. States like Texas, which previously offered generous tax incentives for data centers, are now reconsidering those deals as local communities grapple with the impacts of increased power usage and infrastructure strain.

In Memphis, residents have voiced concerns over SpaceX's data center projects, sparking a wave of public backlash against tech firms. Meanwhile, in other parts of the country, voters are beginning to question whether the economic benefits justify the environmental costs.

These developments underscore a critical tension between innovation and responsibility—between pushing forward with groundbreaking technology and safeguarding the long-term health of our planet and communities.

Looking Forward

The rise of agentic AI is reshaping everything from tech policy to energy consumption. As we move toward a future where AI agents play an even more central role in our lives, we must remain vigilant about the resources being consumed and how they're being managed.

Ultimately, the decisions made today regarding data center infrastructure and power sourcing will determine not only whether AI continues to advance at its current pace—but also whether that advancement remains aligned with sustainable practices and social values.