Introduction to AI Energy Consumption
The energy consumption of AI models has been a topic of interest in recent times. A report by Google sheds light on the energy usage of its Gemini model, a type of AI designed to process and respond to user queries. According to the report, the energy consumption of Gemini prompts can vary greatly, depending on the complexity of the task.
Factors Affecting Energy Consumption
Some Gemini prompts use significantly more energy than others. For instance, feeding dozens of books into Gemini and asking it to produce a detailed synopsis of their content requires more energy than a simple query. Additionally, using a reasoning model can also increase the energy demand, as these models take more steps to produce an answer. The report, however, is limited to text prompts and does not account for the energy required to generate images or videos, which can be much higher.
Energy Consumption Over Time
The total energy used to field a Gemini query has decreased dramatically over time. According to Google, the median Gemini prompt used 33 times more energy in May 2024 than it did in May 2025. This improvement is attributed to advancements in the models and software optimizations. The company estimates that the median prompt uses a relatively small amount of energy, resulting in minimal greenhouse gas emissions.
Greenhouse Gas Emissions
Google estimates the greenhouse gas emissions associated with the median prompt to be approximately 0.03 grams of carbon dioxide. To calculate this, the company multiplies the total energy used to respond to a prompt by the average emissions per unit of electricity. Notably, Google uses a market-based estimate, which takes into account its purchases of clean energy from sources like solar, wind, and geothermal projects. This approach reduces the company’s emissions per unit of electricity to roughly one-third of those on the average grid where it operates.
Water Consumption
In addition to energy consumption, AI data centers also use water for cooling. Google estimates that each prompt consumes about 0.26 milliliters of water, equivalent to approximately five drops. This minimal water usage is a positive aspect of the Gemini model’s environmental impact.
The Goal of Measuring Energy Consumption
The primary goal of this study was to provide users with insight into the energy usage of their interactions with AI. By doing so, Google aims to alleviate concerns about the environmental impact of using AI tools. As stated by Dean, "People are using [AI tools] for all kinds of things, and they shouldn’t have major concerns about the energy usage or the water usage of Gemini models, because in our actual measurements, what we were able to show was that it’s actually equivalent to things you do without even thinking about it on a daily basis."
Conclusion
In conclusion, the energy consumption of AI models like Gemini is a complex topic, influenced by various factors such as prompt complexity and model type. While the energy usage can vary, Google’s report suggests that the median prompt has a relatively low environmental impact. By providing users with a better understanding of AI energy consumption, Google hopes to promote the responsible use of AI tools and mitigate concerns about their environmental effects.
FAQs
Q: How much energy does a Gemini prompt consume?
A: The energy consumption of a Gemini prompt can vary, but the median prompt uses a relatively small amount of energy.
Q: What factors affect the energy consumption of Gemini prompts?
A: The complexity of the task, the type of model used, and the amount of data processed can all impact energy consumption.
Q: How much water does a Gemini prompt consume?
A: Each prompt consumes approximately 0.26 milliliters of water, equivalent to about five drops.
Q: Why does Google use a market-based estimate for emissions?
A: Google uses a market-based estimate to account for its purchases of clean energy from sources like solar and wind power, which reduces its emissions per unit of electricity.
Q: What is the goal of measuring energy consumption in AI models?
A: The goal is to provide users with a better understanding of the environmental impact of AI tools and to promote responsible use.
