In mid-2023, a simple ChatGPT request—like asking for an artichoke pasta recipe—consumed roughly 2 watt-hours of electricity, equivalent to running an incandescent bulb for two minutes.
Fast forward to 2025: OpenAI has launched GPT-5, the most advanced version of its AI chatbot, boasting PhD-level reasoning, website creation, and problem-solving skills.
But experts warn that these upgrades come with a steep environmental price tag.
GPT-5 AI Chatbot Energy Footprint: Up to 20× Higher Than Before
While OpenAI hasn’t officially disclosed GPT-5 AI chatbot exact power usage, independent researchers report that a medium-length response of 1,000 tokens can consume up to 40 watt-hours—about the energy needed to run a light bulb for 40 minutes.
The University of Rhode Island’s AI Lab found GPT-5’s average energy consumption to be around 18 watt-hours per medium response—more than any model they tested except OpenAI’s o3 reasoning model and China’s DeepSeek R1.
Given ChatGPT’s estimated 2.5 billion daily queries, GPT-5’s total power demand could match the daily electricity consumption of 1.5 million U.S. homes.
Why GPT-5 AI Chatbot Uses More Energy
AI experts point to several reasons for GPT-5’s higher consumption:
- Larger Model Size – GPT-5 is believed to be significantly larger than GPT-4, which was already 10× bigger than GPT-3’s 175 billion parameters.
- Reasoning Mode – This feature allows GPT-5 to think through problems in more depth but increases processing time and energy use.
- Multi-Modal Capabilities – GPT-5 can process text, images, and videos, which requires far more computational power than text alone.
- Architecture – While GPT-5 uses a “mixture-of-experts” design to activate only parts of the model per query (improving efficiency), its complexity still raises overall energy demands.
“If you use the reasoning mode, the resources spent for the same answer could be five to ten times higher,” explains Shaolei Ren, Professor at UC Riverside.
Lack of Transparency from AI Companies
OpenAI hasn’t published detailed energy metrics since GPT-3 in 2020.
CEO Sam Altman provided rough figures in June—0.34 watt-hours per query—but without specifying which model they refer to or offering supporting data.
Researchers argue this secrecy is problematic.
“It’s more critical than ever to address AI’s environmental cost,” says Marwan Abdelatti of the University of Rhode Island.
“We call on OpenAI and other developers to publicly disclose GPT-5’s impact.”
The Bigger Picture: AI’s Growing Environmental Impact
A 2024 study by French AI firm Mistral confirmed a strong link between model size and resource consumption—a model 10× bigger can generate 10× more environmental impact for the same amount of text generated.
As the AI industry races toward Artificial General Intelligence (AGI), models like GPT-5 AI chatbot are likely to grow even larger.
Experts caution that without better transparency and efficiency measures, the environmental cost of AI could escalate rapidly.
