OpenAI’s launch of GPT-5 has been met with awe, with its new abilities to create websites and solve complex scientific problems garnering widespread attention. But behind the scenes, a different narrative is emerging. Independent researchers are pulling back the curtain on the model’s resource consumption, and their findings point to a massive increase in energy use. This lack of transparency from the company is prompting a conversation about the true, and often hidden, cost of pushing the boundaries of artificial intelligence.
The numbers are startling. A research group at the University of Rhode Island’s AI lab has found that a medium-length response from GPT-5 can consume an average of 18 watt-hours. This is a significant jump from previous models and is “significantly more energy than GPT-4o,” according to a researcher involved in the study. To visualize this, 18 watt-hours is enough to power an incandescent light bulb for 18 minutes. With billions of queries flowing through ChatGPT daily, the total energy demand could reach the daily electricity needs of 1.5 million US homes, highlighting a sustainability issue of national scale.
A key factor driving this dramatic increase is the model’s size. Though OpenAI has not disclosed the parameter count for GPT-5, experts widely believe it to be “several times larger than GPT-4.” This aligns with a study by the French AI company Mistral, which found a “strong correlation” between a model’s size and its resource consumption. The larger the model, the bigger its environmental impact. This principle, applied to a model believed to be orders of magnitude larger than its predecessors, suggests that the pursuit of ever-more-powerful AI comes with a predictable and escalating cost.
Furthermore, the new capabilities of GPT-5 contribute to its high energy demands. While it employs a “mixture-of-experts” architecture for some efficiency gains, its ability to reason and handle multimodal data like video and images requires more intensive computation. This “reasoning mode” could increase resource usage by a factor of “five to 10” for the same answer, according to one professor. This combination of immense size and complex functionality makes GPT-5 a landmark in both AI capability and environmental impact, and is fueling calls for greater transparency from its creators.