The CEO of OpenAI, Sam Altman, defended the high energy consumption associated with artificial intelligence models during participation in an AI summit at Índia. Ele compared the energy expenditure to train AI systems to that required for human development, highlighting that it takes around 20 years of life and all the food consumed in that period for a person to become intelligent. Altman acknowledged that the public debate about the energy impact of AI is valid, especially considering the accelerated growth in global use of the technology.
The statement was made in a recent interview at the AI Impact Summit event. Altman emphasized that, although model training requires significant amounts of electricity, consumption per individual query is already efficient when compared to equivalent human effort. Ele argued that the main concern should be about the total energy demanded by the expansion of AI in the world.
Criticism of water use in data centers is minimized
Altman called false claims that ChatGPT queries consume large volumes of water, such as 17 gallons per question. Ele explained that older methods of evaporative cooling have been replaced in newer installations that no longer rely on this feature. Alguns modern data centers operate without using water for cooling.
The executive admitted that the topic gained repercussion on the internet, but stated that these estimates do not correspond to current reality. Ele reinforced that the focus should remain on total energy consumption, not on isolated metrics per query.
Data centers represented around 1.5% of global electricity consumption in 2024, according to data from Agência Internacional and Energia. Projeções indicate an annual increase of 15% in this consumption between 2024 and 2030, a rate more than four times higher than that of other sectors.
Transition to clean sources is urgent, says CEO
Altman defended the need to accelerate the adoption of nuclear, wind and solar energy to meet the growing demand of data centers. Ele highlighted that rapid advancement in these sources is essential to sustain the development of AI without compromising environmental goals.
The statement reflects concerns about the current reliance on fossil sources in many projects. The executive argued that AI can contribute to global solutions, but requires adequate energy infrastructure.
Environmental impact of expanding AI
The growth of AI-driven data centers has sparked debates about sustainability. Grupos environmentalists have called for a moratorium on the construction of new facilities in some countries, citing risks to the electricity grid and local communities.
Experts point out that model training consumes most of the energy, while inferences (answers to queries) are more efficient. Altman suggested that fair comparisons consider the total lifetime energy cost of humans versus AI systems.
Challenges for the technology sector
Companies in the sector face pressure to balance innovation and environmental responsibility. OpenAI and other companies invest in partnerships to expand computing capacity, but energy supply remains the main bottleneck.
Altman ruled out solutions like in-orbit data centers as viable at scale this decade. Ele prioritized renewable land sources and nuclear to meet projected demand.
Comparisons generate varied reactions
The analogy between AI training and human development has sparked criticism online. Alguns analysts called the comparison inappropriate, arguing that it ignores ethical and scale differences between individuals and computational models.
Others noted that the focus should be on energy efficiency per task performed. Altman reiterated that AI already outperforms humans in certain aspects of consumption per operation after initial training.
The debate highlights the tension between rapid technological advances and sustainability concerns. OpenAI continues to expand infrastructure while seeking cleaner energy sources.