Altman Wants to Reprice Intelligence: 20 Years of Food Versus One Data Center

Twenty watts. That is how much power the human brain draws while actively working, less than a dim incandescent bulb. For decades, this fact has been the trump card in arguments about the inefficiency of artificial intelligence: why build data centers the size of a small city when nature solved the same problem with one and a half kilograms of neural tissue and a handful of glucose? Sam Altman has decided it is time to revisit the terms of that comparison.

The head of OpenAI has put forward a provocative thesis: comparing the energy consumption of a working brain with that of a working neural network is intellectual sleight of hand. A correct accounting, in Altman's view, must include all the energy spent creating the intelligence, not just running it. For a human, that means roughly twenty years of continuous eating, sleeping and learning: thousands upon thousands of calories converted into synaptic connections. Going further, one would have to count millions of years of evolution, during which countless organisms were born, consumed resources and died so that natural selection could eventually produce a creature capable of reasoning about abstractions.

The argument is not new in academic circles, but coming from the head of the largest AI company, it sounded different: like a strategic narrative. Context is critical here. OpenAI and its competitors face mounting pressure over the energy footprint of large language models.

By various estimates, training a single large model at the GPT-4 level consumed energy comparable to the annual consumption of several thousand households. Each chatbot query uses several times more electricity than an ordinary search query. Building new data centers for AI workloads has become one of the hottest topics in energy policy, from Texas to Scandinavia.

In these conditions, any argument that can reframe the discussion carries not only philosophical but also quite practical weight.

From a scientific standpoint, Altman's position contains a rational kernel, but also noticeable stretches. It is true that if you view the brain as a product of training, its "training budget" is colossal. A child consumes on average between 1,000 and 2,500 kilocalories a day, and a significant share of that energy in the first years of life goes specifically to developing the nervous system.

Over twenty years, this adds up to around 15–18 million kilocalories, roughly 17–21 megawatt-hours when converted to electrical energy. The figure is impressive, but still orders of magnitude smaller than what is spent on training frontier models, once you account for losses to cooling, data transmission and infrastructure. As for evolution, it is technically possible to include it in the calculation, but then AI would also need to carry the entire history of computing, from the first vacuum-tube machines to modern GPU clusters, which makes the comparison meaningless.

Altman's critics were quick to respond. Many researchers pointed to a fundamental difference: the brain is a general-purpose system that simultaneously controls the body, processes sensory information, sustains emotional life and solves intellectual tasks. A large language model does exactly one thing: it generates text (or, in a multimodal variant, images as well). Comparing their "training cost" is like comparing the cost of building an entire city with the cost of building a single skyscraper, however impressive. Other commentators were harsher, calling Altman's remark an attempt to normalize unlimited energy consumption by the AI industry at a moment when society is starting to ask uncomfortable questions.

There is, however, a deeper layer to this discussion. The very fact that the head of OpenAI is publicly reflecting on the price of intelligence signals a shift in industry thinking. Just a couple of years ago, companies preferred not to discuss energy at all, hoping that progress in chip and algorithm efficiency would solve the problem on its own. Now the strategy is changing: instead of denying the scale of consumption, industry leaders are trying to redefine the frame of reference in which it is judged. If intelligence is an expensive product by definition, then high energy costs stop being a bug and become a feature.

This narrative maneuver will have consequences far beyond social media. Regulators in Europe and the US are already drafting energy-efficiency standards for AI systems. How the industry defines the baseline for comparison, whether the cost of a single query, the cost of training a model, or the full cost of creating intelligence "from scratch", will directly affect how strict future rules become. Altman, knowingly or not, is setting the frame for this debate.

One thing can be said for certain: the era when AI companies could simply ignore the question of resources is over. Now they have to build not only models but also arguments that those models are worth the energy spent. And how convincing those arguments turn out to be will determine not only OpenAI's reputation but also the pace of development of the entire industry in the coming years.