OpenAI CEO Sam Altman reveals how much water a single ChatGPT question really uses and it’s not what you think

Artificial intelligence may seem like a clean, invisible power—lines of code producing knowledge at the speed of thought. But what if every question you asked ChatGPT came with a hidden environmental cost? In a recent and unexpected revelation, OpenAI CEO Sam Altman reveals how much water a single ChatGPT question really uses and it’s not what you think.

The truth, as Altman disclosed at a recent conference, has sparked both curiosity and concern within tech circles and beyond. It turns out that AI, especially large-scale models like ChatGPT, isn’t just energy-hungry—it’s thirsty too.

The Shocking Statistic Behind Every ChatGPT Query

Altman’s statement came during a tech and climate panel discussion, where the environmental footprint of artificial intelligence was front and center. “We’ve optimized so much for power efficiency,” he said, “but people don’t realize that AI also consumes a surprising amount of water—particularly during training and even usage.”

The jaw-dropping detail? On average, each ChatGPT interaction—just one prompt and response—can indirectly use a few hundred milliliters of water, especially when processed in certain data centers that rely on water-based cooling systems.

So while the interaction may seem like a simple question typed on your screen, its carbon and water cost is part of a larger infrastructure story.

Why Does ChatGPT Use Water at All?

You may be wondering: why would AI, which runs in data centers powered by electricity, use water at all?

The answer lies in how data centers manage heat. Large-scale computing infrastructure generates enormous amounts of heat, especially during the training of massive models like GPT-4 and in real-time inference when millions of people use the model daily.

To keep these machines cool, companies use liquid-cooling systems. Many data centers draw on local water sources—sometimes from municipal systems, sometimes from groundwater or industrial supplies—to absorb and remove this heat efficiently.

Altman clarified that this cooling is necessary to prevent hardware degradation and service interruptions. But it also represents an unseen environmental impact of modern AI.

Training vs. Using: Where the Water Goes

Training a model like GPT-4, according to studies cited by Altman and confirmed by third-party researchers, can consume millions of liters of water. That’s because training is a one-time, extremely compute-heavy process that can span weeks or months, depending on the model's complexity.

In contrast, inference—or daily use—has a smaller but continuous water footprint. Every prompt you enter into ChatGPT activates servers that must stay cool, 24/7, across a network of global data centers.

Altman noted, “The cost per question may seem small—less than a bottle of water—but when scaled to billions of queries per day, it adds up.”

Public Reaction: A Mix of Surprise and Concern

The public response to Altman’s statement was swift. Some were shocked: “I never thought I’d be thinking about water use when chatting with an AI,” said one Reddit user.

Others called for greater transparency from AI companies:“If we know the carbon footprint of our actions, why not the water footprint too?” asked an environmental journalist on X (formerly Twitter).

The revelation has sparked new questions about sustainable AI practices, especially as the tech industry faces increasing pressure to address climate change and resource depletion.

What OpenAI Is Doing About It

To his credit, Altman didn’t just drop the fact and move on. He acknowledged the issue and pointed to OpenAI’s ongoing efforts to minimize water and energy usage across its operations.

“We are actively exploring more sustainable data center partnerships, including those that use recycled or non-potable water for cooling,” he said. “We’re also pushing for more efficient model architectures that require fewer resources during both training and usage.”

OpenAI is also participating in broader industry conversations about AI’s environmental impact, advocating for standardized reporting metrics for energy and water use in AI systems.

The Bigger Picture: AI’s Growing Environmental Footprint

AI is now at the heart of everything—from healthcare diagnostics to education and customer service. But the infrastructure powering it consumes not just electricity, but also natural resources like water.

With estimates suggesting that data centers will consume up to 10% of global electricity by 2030, their accompanying water use must be taken seriously.

This is especially pressing in regions already suffering from water stress, where data center operations could strain local resources.

Altman’s transparency adds urgency to the need for climate-responsible AI—and opens up discussions for future technologies like air-cooled or geothermal data centers, as well as edge computing that reduces reliance on massive centralized hubs.

Should You Stop Using ChatGPT?

Not necessarily.

While the average person’s use of ChatGPT isn’t likely to cause environmental harm on its own, Altman’s revelation reminds us of a broader truth: every digital action has a real-world cost.

What’s important now is awareness, transparency, and industry-wide responsibility. Just as we’ve learned to unplug devices, reduce streaming resolution, or drive less for sustainability, we may one day have AI-specific best practices—powered not just by innovation, but by intention.

Disclaimer

The information provided on GCCNews24 (GCCNews24.com) is for general informational purposes only. While we strive to ensure that the news, opinions, and other content published are accurate and up-to-date, GCCNews24 makes no representations or warranties of any kind, express or implied, regarding the completeness, accuracy, reliability, or suitability of the information contained on the website for any purpose.