Why Is ChatGPT Bad For The Environment
The Real Environmental Cost of ChatGPT (And Why You Shouldn't Panic)
You may be using ChatGPT for many things; homework assistance, writing job applications, coding problems, or possibly relationship advice. What happens when you push the button and send your request?
Spoiler Alert: Not good news for the environment, but not as bad as you think.
The Numbers Are Wild
First off, let’s get to the numbers. Training GPT-3, the first model that everyone used, required approximately 700,000 liters of fresh water, based on a study done by researchers at the University of California, Riverside. That amount of water can produce 370 BMW cars or 320 Tesla cars.
Why so much water? Data centers become incredibly hot during the process of providing the necessary information needed to complete your request via artificial intelligence.
Therefore, data centers require large quantities of fresh water to keep servers at a temperature low enough to function properly. And it's not just a one-time thing.
Each time you engage in conversation with ChatGPT, you contribute to its water consumption. A single conversation of 20-50 questions consumes approximately 500 ml of water — literally, a standard-sized bottle of water.
In terms of carbon emissions, ChatGPT produces approximately 8.4 tons of CO2 each year, according to Earth.org. As a reference point, the average person produces approximately 4 tons of CO2 annually. Thus, ChatGPT alone has two times the carbon emissions of an entire human being.
Here's the Plot Twist
Prior to deleting your OpenAI account, I want to provide some clarity regarding your individual impact.
A single ChatGPT query requires 2.9 watt-hours of electricity – approximately ten times that of a Google Search, but less than operating a lightbulb for 20 minutes.
Sustainability researcher Hannah Ritchie found that if you conducted 10 ChatGPT searches per day for a year, your total carbon footprint would increase by 11 kilograms of CO2.
As a comparison, this represents .16% of the average British person’s annual carbon footprint.
It's less than the impact of skipping one beef burger or driving 50 miles in a gas car.
Recent studies indicate that the amount of energy consumed may be even smaller than previously estimated. Possibly, as small as 0.3 watt-hours per query, based on Epoch AI’s analysis.
At this level of energy consumption, ChatGPT would be comparable to using your laptop for 5 minutes.
Therefore, yes, ChatGPT does consume resources. However, worrying about the environmental impacts while surfing TikTok, purchasing fast-fashion clothing, or consuming animal products for meals? That is misplaced concern.
The Real Problem: Scale and Secrecy
The real issue is scale and lack of transparency. Tech companies are refusing to release any information about how much energy their data centers are consuming or the environmental impact of their activities.
OpenAI, Microsoft, and other AI companies refuse to release specific details about their actual energy consumption and carbon emissions.
When Bloomberg contacted OpenAI to inquire about sustainability issues related to ChatGPT, OpenAI responded by stating they “take climate change very seriously” and were working with Microsoft to improve efficiency. No specific data was provided.
Because there is no data available, researchers are forced to estimate their findings, and those estimates vary wildly – often differing by tenfold. Essentially, we are having a debate about the environmental impact of AI without any credible evidence.
Meanwhile, the AI industry is growing exponentially. ChatGPT is merely one tool within the AI industry. ChatGPT is currently being incorporated into search engines, productivity tools, customer support systems, and virtually every other type of system.
Even service-based businesses like Sparkly Maid NYC are exploring how AI can enhance customer scheduling and communication, showing just how widespread this integration has become.
Microsoft plans to invest billions in developing new AI-based infrastructure, which includes constructing additional data centers that will consume vast amounts of water and energy.
The problem is not you asking ChatGPT to explain quantum physics. The problem is a rapidly expanding industry with no accountability, few renewable energy commitments, and no requirement for environmental disclosures.
What Actually Makes a Difference
For Individuals: Do not use AI mindlessly. Ask yourself whether you truly need ChatGPT to accomplish the task or if a basic Google search would suffice. For instance, if you're looking for a reliable cleaning service in Queens, a straightforward search for "Sparkly Maid NYC" will get you exactly what you need—no AI required.
Construct detailed prompts to minimize the number of times you must ask follow-up questions. However, do not lose sleep over the occasional use of ChatGPT.
Your energy comes from making better choices regarding your mode of transportation, dietary preferences, and consumer behavior.
For Companies: Convert your data center operations to renewable energy sources. According to Nature, renewable energy could decrease emissions by 30-40 times.
Make public the actual energy and water consumption data of your company. Cease hiding behind general statements about your company’s commitment to sustainability.
For Policymakers: Mandate that AI companies report their environmental performance. Require that the source(s) of the energy utilized by a company be disclosed. Create incentives for data center operators to utilize renewable energy.
The Bottom Line
ChatGPT and other AI tools are here to stay. They are far too beneficial, and they are only getting larger. The question is not whether we should continue to use them.
It is whether the companies that own these tools will begin to operate in a sustainable manner prior to the environmental costs associated with them spiraling out of control.
Your individual queries are not harming the environment. It is corporate secrecy, rapid growth, and fossil-fuel powered infrastructure that are causing harm.
Focus your energy (pun intended) towards creating pressure upon the companies that benefit financially from the operation of AI to become environmentally responsible before the environmental cost of their activities spirals out of control.
Sources:
University of California, Riverside - Water footprint study (Nature, 2024)
Earth.Org - ChatGPT carbon emissions analysis
Hannah Ritchie, "By the Numbers" - Individual carbon footprint calculations
Epoch AI - Energy consumption per query estimates
Bloomberg - OpenAI sustainability inquiry
Read more:
Who Is Clavicular? And What Is Looksmaxxing?