AI’s integration into our lives is the most significant shift in online life in more than a decade. Hundreds of millions of people now regularly turn to chatbots for help with homework, research, coding, or to create images and videos. But what’s powering all of that?

Today, new analysis by MIT Technology Review provides an unprecedented and comprehensive look at how much energy the AI industry uses—down to a single query—to trace where its carbon footprint stands now, and where it’s headed, as AI barrels towards billions of daily users.

This story is a part of MIT Technology Review’s series “Power Hungry: AI and our energy future,” on the energy demands and carbon costs of the artificial-intelligence revolution.

We spoke to two dozen experts measuring AI’s energy demands, evaluated different AI models and prompts, pored over hundreds of pages of projections and reports, and questioned top AI model makers about their plans. Ultimately, we found that the common understanding of AI’s energy consumption is full of holes.

We started small, as the question of how much a single query costs is vitally important to understanding the bigger picture. That’s because those queries are being built into ever more applications beyond standalone chatbots: from search, to agents, to the mundane daily apps we use to track our fitness, shop online, or book a flight. The energy resources required to power this artificial-intelligence revolution are staggering, and the world’s biggest tech companies have made it a top priority to harness ever more of that energy, aiming to reshape our energy grids in the process.

Meta and Microsoft are working to fire up new nuclear power plants. OpenAI and President Donald Trump announced the Stargate initiative, which aims to spend $500 billion—more than the Apollo space program—to build as many as 10 data centers (each of which could require five gigawatts, more than the total power demand from the state of New Hampshire). Apple announced plans to spend $500 billion on manufacturing and data centers in the US over the next four years. Google expects to spend $75 billion on AI infrastructure alone in 2025.

This isn’t simply the norm of a digital world. It’s unique to AI, and a marked departure from Big Tech’s electricity appetite in the recent past. From 2005 to 2017, the amount of electricity going to data centers remained quite flat thanks to increases in efficiency, despite the construction of armies of new data centers to serve the rise of cloud-based online services, from Facebook to Netflix. In 2017, AI began to change everything. Data centers started getting built with energy-intensive hardware designed for AI, which led them to double their electricity consumption by 2023. The latest reports show that 4.4% of all the energy in the US now goes toward data centers.

The carbon intensity of electricity used by data centers was 48% higher than the US average.

Given the direction AI is headed—more personalized, able to reason and solve complex problems on our behalf, and everywhere we look—it’s likely that our AI footprint today is the smallest it will ever be. According to new projections published by Lawrence Berkeley National Laboratory in December, by 2028 more than half of the electricity going to data centers will be used for AI. At that point, AI alone could consume as much electricity annually as 22% of all US households.

Meanwhile, data centers are expected to continue trending toward using dirtier, more carbon-intensive forms of energy (like gas) to fill immediate needs, leaving clouds of emissions in their wake. And all of this growth is for a new technology that’s still finding its footing, and in many applications—education, medical advice, legal analysis—might be the wrong tool for the job or at least have a less energy-intensive alternative.

Tallies of AI’s energy use often short-circuit the conversation—either by scolding individual behavior, or by triggering comparisons to bigger climate offenders. Both reactions dodge the point: AI is unavoidable, and even if a single query is low-impact, governments and companies are now shaping a much larger energy future around AI’s needs.

We’re taking a different approach with an accounting meant to inform the many decisions still ahead: where data centers go, what powers them, and how to make the growing toll of AI visible and accountable.

ChatGPT is now estimated to be the fifth-most visited website in the world, just after Instagram and ahead of X.

That’s because despite the ambitious AI vision set forth by tech companies, utility providers, and the federal government, details of how this future might come about are murky. Scientists, federally funded research facilities, activists, and energy companies argue that leading AI companies and data center operators disclose too little about their activities. Companies building and deploying AI models are largely quiet when it comes to answering a central question: Just how much energy does interacting with one of these models use? And what sorts of energy sources will power AI’s future?

This leaves even those whose job it is to predict energy demands forced to assemble a puzzle with countless missing pieces, making it nearly impossible to plan for AI’s future impact on energy grids and emissions. Worse, the deals that utility companies make with the data centers will likely transfer the costs of the AI revolution to the rest of us, in the form of higher electricity bills.

It’s a lot to take in. To describe the big picture of what that future looks like, we have to start at the beginning.

Before you can ask an AI model to help you with travel plans or generate a video, the model is born in a data center.

Racks of servers hum along for months, ingesting training data, crunching numbers, and performing computations. This is a time-consuming and expensive process—it’s estimated that training OpenAI’s GPT-4 took over $100 million and consumed 50 gigawatt-hours of energy, enough to power San Francisco for three days. It’s only after this training, when consumers or customers “inference” the AI models to get answers or generate outputs, that model makers hope to recoup their massive costs and eventually turn a profit.