Every ChatGPT Question Comes at a Cost: Enough Power for a Lightbulb and a Teaspoon of Water

In an increasingly digital world, the convenience of artificial intelligence like ChatGPT has become commonplace.

Yet, behind each seemingly effortless query lies a hidden environmental toll.

From the energy powering vast data centres to the surprising amount of water used for cooling, the true cost of our AI interactions extends far beyond a simple internet connection.

OpenAI CEO Sam Altman has now shed more light on the environmental footprint of AI tools like ChatGPT.

In a recent blog post, he revealed that an average ChatGPT query uses just 0.000085 gallons of water, which he described as 'roughly one-fifteenth of a teaspoon.'

The Environmental Cost of ChatGPT

This insight comes amidst growing concerns over how much energy and water artificial intelligence systems consume. Altman's recent blog post, while discussing the future of AI, also presented specific figures regarding ChatGPT's resource consumption.

'People are often curious about how much energy a ChatGPT query uses; the average query uses about 0.34 watt-hours, about what an oven would use in a little over one second, or a high-efficiency lightbulb would use in a couple of minutes,' Altman wrote.

'It also uses about 0.000085 gallons of water, roughly one-fifteenth of a teaspoon,' he added. The water usage of one-fifteenth of a teaspoon per query might appear tiny for one person. Yet, given the billions of queries AI systems handle daily, the total effect becomes substantial. OpenAI did not explain how these figures were arrived at.

'As data centre production gets automated, the cost of intelligence should eventually converge to near the cost of electricity.' Altman continued. However, this perspective on electricity cost also underscores the larger environmental concerns now under close examination.

Increased Focus on AI's Environmental Burden

As AI becomes more widely adopted, experts and researchers are concerned about its environmental burden. Studies earlier this year forecast that AI could potentially use more electricity than Bitcoin mining by late 2025.

Water consumption also presents a major worry, particularly for water-based cooling data centres. According to a 2024 report from The Washington Post, creating a 100-word email with GPT-4 might use roughly one bottle of water, influenced by the data centre's site and cooling approach. This demonstrates how environmental effects can differ depending on the infrastructure.

Altman suggests that, in time, the expense of producing intelligence via AI will nearly match that of electricity. Until then, discussions concerning AI's environmental impact will become more vocal.

Every Drop Counts: AI's Thirst for Water

This isn't the only occasion Altman has foreseen AI becoming more affordable to run. In February, Altman stated in a blog post that AI's operating expenses would reduce tenfold yearly.

'You can see this in the token cost from GPT-4 in early 2023 to GPT-4o in mid-2024, where the price per token dropped about 150x in that time period,' Altman wrote. 'Moore's law changed the world at 2x every 18 months; this is unbelievably stronger,' he added.

Big Tech's Ambitious Energy Drive

Leading tech firms in the AI race are exploring nuclear energy as a power source for their data centres. In September, Microsoft finalised a 20-year agreement with Constellation Energy to restart a dormant nuclear plant at Three Mile Island.

In October, Google revealed it had partnered with Kairos Power, a nuclear energy company, to produce three small modular nuclear reactors. These reactors, which can supply up to 500 megawatts of electricity, are expected to be operational by 2035.

In an October interview, Google's CEO, Sundar Pichai, told Nikkei Asia that the search giant aims for net-zero emissions across its entire operations by 2030. He further stated that Google also assessed solar energy beyond just nuclear power.

'It was a very ambitious target, and we are still going to be working very ambitiously towards it. Obviously, the trajectory of AI investments has added to the scale of the task needed,' Pichai said.