Altman: Humans guzzle energy just like AI - AI News Today Recency

# Altman: Humans Guzzle Energy Just Like AI

OpenAI CEO Sam Altman has sparked renewed debate about artificial intelligence's environmental footprint by arguing that the energy consumption of AI systems should be compared fairly to human energy usage rather than analyzed in isolation[2][3]. Speaking at the India AI Impact Summit this week, Altman challenged what he views as misleading narratives surrounding AI's power demands, contending that when measured on an equivalent basis, artificial intelligence has already achieved energy efficiency parity with human cognitive work[2][3].

The Energy Debate: AI Versus Human Consumption

Altman's comments come amid growing scrutiny of AI's environmental impact, particularly as data centers consume an ever-larger share of global electricity[4]. An average ChatGPT query uses approximately 0.34 watt-hours of electricity—equivalent to what an oven consumes in just over one second or an LED bulb in a couple of minutes[1][4]. While these individual figures appear negligible, the cumulative effect becomes significant when multiplied across millions of daily queries[4].

However, Altman argues this metric misses the broader context. He contends that fair comparisons should measure how much energy it takes for AI to answer a question once its model is trained versus how much energy a human requires to perform the same cognitive task[2]. "Probably, AI has already caught up on an energy efficiency basis, measured that way," Altman stated[2]. To illustrate this point, he drew a parallel to human development, noting that training a human takes approximately 20 years of food consumption[5].

Disputing Water Usage Claims and "Unfair" Comparisons

Altman dismissed concerns about ChatGPT's water consumption as "totally fake" and "completely untrue, totally insane, no connection to reality"[2]. He acknowledged that water usage was a legitimate concern when data centers used evaporative cooling, but stressed that modern facilities no longer employ this method[2]. The OpenAI CEO criticized what he views as misleading claims circulating online suggesting ChatGPT uses 17 gallons of water per query[2].

The CEO also objected to comparisons that measure energy consumption per training instance rather than per inference query[2]. He argued these comparisons are fundamentally unfair because they conflate the one-time energy cost of training a model with the ongoing operational energy required to answer individual questions[2]. When adjusted for this distinction, Altman maintains that AI systems demonstrate superior energy efficiency compared to human alternatives[3].

The Real Challenge: Total Energy Consumption and Future Solutions

While defending AI's per-query efficiency, Altman acknowledged that total energy consumption represents a legitimate concern[2]. "It's fair to be concerned about the energy consumption — not per query, but in total, because the world is now using so much AI," he stated[2]. Data centers currently consume 4–5% of all electricity in the United States, with that share climbing rapidly as AI workloads expand[4]. According to a Department of Energy–funded report, data center energy consumption more than doubled from 2017 to 2023, reaching approximately 176 terawatt-hours[4].

To address this challenge, Altman advocates for accelerated transition toward renewable and nuclear energy sources[2]. "The world needs to move towards nuclear or wind and solar very quickly," he said[2]. A recent study by the University of California, Riverside and the Washington Post found that ChatGPT alone uses nearly 40 million kilowatts of energy per day—enough to power the Empire State Building for approximately 18 months[1].

Frequently Asked Questions

How much electricity does a single ChatGPT query consume?

An average ChatGPT query uses approximately 0.34 watt-hours of electricity, equivalent to the energy an oven consumes in just over one second or an LED bulb uses in a couple of minutes[1][4].

Is ChatGPT's water usage a significant environmental concern?

According to Altman, current water usage concerns are exaggerated because modern data centers no longer use evaporative cooling methods[2]. However, ChatGPT does use approximately 0.000085 gallons of water per query, roughly one-fifteenth of a teaspoon[1][4].

How does AI's energy efficiency compare to human cognitive work?

Altman argues that when measured fairly—comparing the energy required for AI to answer a question after training versus the energy a human needs to perform equivalent cognitive work—AI has already achieved energy efficiency parity with humans[2][3].

What percentage of U.S. electricity do data centers consume?

Data centers currently consume approximately 4–5% of all electricity in the United States, with this share climbing as AI workloads increase[4]. Data center energy consumption more than doubled from 2017 to 2023, reaching approximately 176 terawatt-hours[4].

What is the total daily energy consumption of ChatGPT?

ChatGPT uses nearly 40 million kilowatts of energy per day, sufficient to power the Empire State Building for approximately 18 months or charge eight million smartphones[1].

What solutions does Altman propose to address AI's energy demands?

Altman advocates for rapid transition toward renewable energy sources, specifically nuclear, wind, and solar power, to support the growing energy demands of AI systems[2].