OpenAI GPT-5 Energy Use: Hidden Costs & Potential Increase
# GPT-5’s Environmental Footprint: What We Know About the Energy Cost of OpenAI’s Latest Model
The release of OpenAI’s GPT-5 has sparked excitement about the future of artificial intelligence, but also renewed concerns about its environmental impact. Larger AI models require important computational resources, translating to ample energy consumption. While OpenAI hasn’t publicly disclosed detailed data about GPT-5’s energy usage, researchers are beginning to piece together an understanding of its potential footprint – and the need for greater transparency from AI developers.## The Scaling Problem: Bigger Models, Bigger Energy Bills
A key concern is the relationship between model size and energy consumption. according to researchers at the University of Rhode Island, a 10x increase in model size can lead to a one-order-of-magnitude increase in energy usage for the same amount of generated text. This means that as models like GPT-5 grow exponentially in complexity, their energy demands also increase dramatically.
Though, scale isn’t the only factor. Researchers Jegham, Kumar, and Ren point out that GPT-5 benefits from deployment on more efficient hardware compared to previous generations. Moreover, its architecture likely incorporates a “mixture-of-experts” approach. This streamlined design activates only a portion of the model’s parameters for each query, potentially reducing energy consumption.
Despite these efficiencies, GPT-5’s expanded capabilities - particularly its ability to process video, images, *and* text – likely offset some of the gains.The addition of reasoning capabilities also plays a role.”If you use the reasoning mode, the amount of resources you spend for getting the same answer will likely be several times higher, five to 10,” explains Ren. This is because reasoning requires more computational steps and longer processing times.
## How Researchers Are Estimating GPT-5’s Energy Use
Calculating the resource consumption of AI models is a complex undertaking. The University of rhode Island team tackled this challenge by multiplying the average response time of a model by its average power draw during operation.
Estimating power draw proved particularly challenging. “It was a lot of work,” says Abdeltawab Hendawi, a professor of data science at the University of Rhode Island. The team faced challenges in obtaining information about how AI models are deployed within data centers, and how queries are distributed across different chips.Their research, detailed in a recent paper,includes estimates for the chips used and the query distribution process.
Interestingly, their findings align with data released by OpenAI.Altman’s June blog post reported ChatGPT’s energy consumption at 0.34 watt-hours per query - a figure that closely matches the team’s estimates for GPT-4o. This validation underscores the accuracy of their methodology.
### The Importance of transparency
The research highlights a critical need for greater transparency from AI companies. as models continue to grow in size and capability, understanding their environmental impact becomes increasingly crucial. Without access to detailed information, it’s difficult to assess the true cost of AI progress and deployment.
“It’s more critical than ever to address AI’s true environmental cost,” emphasizes Marwan Abdelatti, a professor at URI. “We call on OpenAI and othre developers to commit to full transparency by publicly disclosing GPT-5’s environmental impact.”
This call for transparency isn’t just about environmental duty; it’s also about fostering public trust and enabling informed discussions about the future of AI.OpenAI’s decision to share data on ChatGPT’s energy consumption is a positive step, and researchers hope they will continue to prioritize transparency as they release increasingly powerful models like GPT-5. The future of AI depends not only on innovation, but also on a commitment to sustainability and responsible development.