ChatGPT is easy to use, but each of your inquiries carries an imperceptible environmental cost. OpenAI CEO Sam Altman recently lifted the veil on just what that cost is, and the figures expose both the magnitude of AI’s footprint today and the difficulty of the road ahead.
As Altman estimated in his June 11 blog post “The Gentle Singularity,” every ChatGPT question consumes about 0.34 watt-hours of energy. For perspective, that’s comparable to having your oven on for a little over a second or a high-efficiency lightbulb burning for a few minutes. That doesn’t sound like much, but that’s where the math really gets exciting.
ChatGPT processes hundreds of millions of requests each day. When you multiply those seemingly tiny energy costs across that massive user base, the environmental footprint starts looking a lot less tiny. It’s the classic problem of scale that defines our digital age: individual actions feel negligible, but collective impact tells a different story.
The Hidden Water Cost
Energy consumption isn’t the only resource at stake. Altman also revealed that each query uses approximately 0.000085 gallons of water, roughly one-fifteenth of a teaspoon. This water is largely dedicated to cooling enormous data centers in which AI technology operates, and these data centers produce colossal quantities of heat when running.
Again, that teaspoon amount sounds ridiculously tiny until you consider the volume. Hundreds of millions of users make several queries per user adds up to a lot of water usage, particularly in the context of the reality that most data centers are located in environments already short on water.
These sources emerge as experts warn of the increasing thirst for resources by AI. Others predict that, soon enough, artificial intelligence will equal or surpass the energy consumption of cryptocurrency mining, and cryptocurrency mining has already been widely criticized for its ecological footprint.
The Promise and the Problem of ChatGPT
Altman isn’t just acknowledging the current costs; he’s painting a picture of a radically different future. He predicts that by the 2030s, both intelligence and energy will become “wildly abundant.” This abundance, he suggests, could unlock breakthrough discoveries in physics, enable space colonization, and advance brain-computer interfaces beyond what we can currently imagine.
“With abundant intelligence and energy (and good governance), we can theoretically have anything else,” Altman wrote, presenting an optimistic vision where AI helps solve the very environmental challenges it currently contributes to.
But he’s also realistic about the risks. When a system serves hundreds of millions of people daily, even small problems can spiral into massive consequences. “A small new capability can create a hugely positive impact; a small misalignment multiplied by hundreds of millions of people can cause a great deal of negative impact,” he cautioned.
The Economics of AI: Evolution of ChatGPT
One factor working in favor of sustainability is the rapid decline in AI costs. Altman pointed to the dramatic price drop between GPT-4’s early 2023 launch and GPT-4o in mid-2024, where costs per token plummeted by about 150 times. He predicts this trend will continue, with AI costs potentially dropping tenfold each year.
This isn’t just good news for consumers and businesses; it could also mean better environmental efficiency. As the technology becomes cheaper to run, it often becomes more energy-efficient too. The challenge lies in whether efficiency improvements can outpace the explosive growth in usage.
What This Means Going Forward
The case for AI’s impact on the environment is only just starting. As the machines grow stronger and more deeply integrated into our daily lives, the strain on resources can only get stronger. The question isn’t whether or not AI will affect the environment; it’s how we can do it in a sustainable way without depriving ourselves of the advantages of the technology.
Altman’s disclosure of these costs and investments is a step in the right direction. Quantifying AI’s resource usage is allowing us to begin having reasonable discussions about tradeoffs, optimization, and ethical development practices. The future he describes, where clean energy is plentiful and so is intelligence, is appealing. The way we get to it ethically will depend on the kind of open accounting he’s started to offer.
