Each time you ask ChatGPT a question, it just doesn’t give you an answer—it also draws energy, water, and materials. The responses are thus smooth, but the environmental impact is huge, and many users are not aware of it.

The Energy Guzzler in the Cloud

OpenAI’s ChatGPT, together with other large language models (LLMs), operates on strong data centers packed with lots of servers. These servers have to be supplied with electricity to a great extent to perform requests, train models, and run AI continuously without any interruptions.

A study shows that the training of one AI model, such as GPT-3 GPT-3, takes up as much power as 120 families. And it is only the initial part of the training; every new work adds to the energy consumption.

Water: The Overlooked Resource

Data centers also need water, apart from the electricity, which is used in the process of cooling. One data center may use a huge amount of water to cool down computers.

Researchers say that ten to fifty answers from ChatGPT might need the same amount of water that half a liter represents, though this depends on where the server is located and what kinds of cooling are used. Water scarcity for irrigation areas is serious; therefore, it cannot be neglected when developing sustainable water management plans.

The Carbon Footprint of AI Conversations

Additionally, it is observed that carbon emissions due to AI usage may cause air pollution to become a matter of growing concern. Although these firms invest in green energy, there are still many places that use fossil fuel as the main power source.

In a study conducted by researchers at the University of Massachusetts, it was discovered that creating and training an AI model can result in the release of 626,000 pounds of carbon dioxide into the atmosphere, which is equivalent to the emissions of five average cars over their whole lives. Additionally, even small daily ChatGPT queries add to this carbon footprint, albeit to a lesser extent.

Can AI Ever Be Sustainable?

Tech companies are looking for ways to lessen AI’s impact on the environment. Here are some of their plans:

• More efficient algorithms that need less computing power.
• Powering data centers with renewables to reduce dependence on fossil fuels.
• Water-efficient cooling methods, for example, liquid immersion cooling.

Nevertheless, as AI becomes more capable and is employed in many areas, its appetite for energy is likely to keep growing.

What Can Users Do?

Consumers, on the other hand, cannot fully eliminate environmental harm caused by AI, but their awareness can be helpful:

• Not asking unnecessary questions—remember that each interaction is not free.
• Using the services of companies that are concerned with the environment.
• Requesting access to sustainability reports from technology companies.

The Bottom Line

The electric bill for ChatGPT’s usefulness is usually concealed. The more AI is embraced in our lives, the more we need to find ways to use it sustainably. That is why, before you put a query to ChatGPT, think that something may be hidden.