CEO Compares Humans and AI, Says ‘Training a Human Takes 20 Years of Food’ Before They Get Smart

When Sam Altman said it takes 20 years of food to make a human smart, the internet did not take it well. OpenAI CEO Sam Altman has never been shy about defending artificial intelligence. But a remark he made at the India AI Impact Summit cut differently. Speaking to The Indian Express, Altman pushed back against critics of AI’s energy footprint by flipping the conversation entirely and pointing the finger at human biology itself.

The comment came during a wide-ranging session where Altman was asked about the environmental cost of running systems like ChatGPT. He called direct comparisons between AI training costs and a single human query “always unfair.” What followed was the line that set the internet on fire.

“It also takes a lot of energy to train a human,” Altman said. “It takes like 20 years of life and all of the food you eat during that time before you get smart. And not only that, it took the very widespread evolution of the 100 billion people that have ever lived and learned not to get eaten by predators and learned how to figure out science and whatever, to produce you.” The audience laughed. Much of the internet did not.

What Altman Was Actually Arguing

Behind the provocative framing was a specific technical point, but critics say it leaves out most of the story. Altman told The Indian Express that critics often use the wrong math. He argued that if people want a fair comparison, they should think about how much energy it takes to create a human expert in the first place, rather than comparing AI training to a single human query. In his view, once a model is trained, each individual response becomes highly efficient.

He framed it this way: the fair comparison is how much energy a trained AI model uses to answer a question versus how much a human uses to do the same, and by that measure, “probably AI has already caught up on an energy efficiency basis.” He also cited estimates placing a single ChatGPT query at roughly 0.34 watt-hours of electricity, about the same as running an oven for one second.

Altman went further on water usage, brushing off viral claims that each ChatGPT query consumes gallons of water. “Don’t use ChatGPT, it’s 17 gallons of water for each query or whatever,” he said, mimicking the online claims. “This is completely untrue, totally insane, no connection to reality.” He attributed the old figures to evaporative cooling systems that modern data centers have largely moved away from.

What the Data Actually Shows

Altman’s argument got pushback, not just online, but from researchers actively tracking AI’s environmental footprint. According to the Environmental and Energy Study Institute, large data centers can consume up to five million gallons of water per day, equivalent to the water use of a town with up to 50,000 people. The institute also noted that as new AI-focused facilities come online, water consumption is climbing alongside energy use and carbon emissions.

A January report from water technology company Xylem and Global Water Intelligence found that 56% of data centers globally still rely on evaporative cooling in some form, casting doubt on Altman’s suggestion that the water concern is largely a relic of outdated infrastructure. The picture on electricity is similarly hard to dismiss.

A US Department of Energy report estimates that data center energy usage in the United States alone could reach as much as 580 TWh by 2028, equivalent to around 12% of the nation’s total electricity consumption. The International Energy Agency projects that by 2030, global data center electricity consumption will more than double, reaching 945 TWh, slightly exceeding Japan’s entire electricity use today.

The Backlash and What It Reveals

The numbers are one thing. The reaction to how Altman framed the argument is a separate conversation entirely. Critics argued that what sounded like a hastily crafted metaphor exposed a deeper disconnect between some AI leadership and the public, one that frames human existence and resource use in the same utilitarian terms as server racks and silicon. The remark spread quickly, and not in Altman’s favor.

Futurism noted that OpenAI, Microsoft, and Amazon do not publicly disclose their energy bills, CO2 emissions, or water consumption tied to AI, making it difficult for the public to verify the efficiency claims Altman was making on a public stage. Transparency, in that sense, remains a one-way argument.

Altman’s insistence that humans and AI are comparable in energy consumption is symptomatic of a broader reckoning the tech industry must face: intelligence, whether biological or artificial, cannot be reduced to calories burned or electricity consumed without losing sight of what makes human life meaningful. For many observers, that is exactly the kind of framing that makes these conversations so difficult.

Where Altman Stands and Where the Debate Goes Next

To be fair, Altman did not deny that AI consumes significant power overall. “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,” Altman said, adding that the solution is to “move towards nuclear or wind and solar very quickly.” He described the global AI infrastructure buildout as potentially the most expensive and complex infrastructure project humanity has ever taken on.

The comments aligned with OpenAI’s broader push for advanced AI amid competition from Google, Anthropic, Meta, and others, with Altman stressing that breakthroughs in energy production, including nuclear power, are necessary to sustain progress. The argument is not that AI has no cost. It is that the cost is worth paying, and that humans are expensive too.

Whether that argument lands depends on who you ask. TechCrunch and Cybernews both covered the backlash extensively, and neither framed it as a winning moment for Altman. The deeper question his comments raised, about who gets to define what counts as a fair use of energy and who answers for the environmental cost of building the AI economy, is one the industry has yet to answer clearly.