Your ChatGPT query is killing nature

Every time you pose a question to ChatGPT, it uses approximately 0.0029 kilowatt-hours of electricity, which is almost 10X the amount of energy required for a standard Google search Image: Shutterstock

ChatGPT hit an estimated 100 million monthly users within just two months of its launch in 2022. While many see it as a useful tool which eases human effort, others see it slowly taking over jobs, and making people less creative and productive. But that isn’t the only concern that the advent of ChatGPT has brought forward. For every help that the artificial intelligence is providing a human, it is also taking a lot from the environment. As per a recent report by a US-based research platform BestBrokers.com, for answering questions alone, ChatGPT consumes a whopping 226.82 million kilowatt-hours on average every year, which amounts to an expenditure of $29.71 million.

 ChatGPT’s Energy Consumption for responding to prompts and its cost in the US

Every time you pose a question to ChatGPT, it uses approximately 0.0029 kilowatt-hours of electricity, which is almost 10X the amount of energy required for a standard Google search, which uses around 0.0003 kilowatt-hours per search, as reported by The Electric Power Research Institute (EPRI).

Also read: How will AI reshape our world? It's really up to us

How much energy does ChatGPT consume every week?

As of November 2023, OpenAI CEO Sam Altman reports that 100 million users engage with ChatGPT weekly, generating about 1.5 billion queries. This equates to over 214 million daily requests, consuming more than 500,000 kilowatt-hours of energy each day. For comparison, the average US household uses around 29 kilowatt-hours daily, meaning ChatGPT uses over 21,600 times more power than a typical household each day.

And in a year?

Annually, ChatGPT consumes 226,821,615 kilowatt-hours of energy, handling around 78 billion queries. At $0.131 per kWh, this costs about $29.7 million, with each query costing $0.00038. Upgrades, new features could increase these figures.

Also read: WHO weighs up AI risks and benefits for healthcare

If used elsewhere…

The annual amount of electricity ChatGPT uses to respond to prompts (226.82 million kWh) would be enough to:

• Fully charge 3.13 million electric vehicles once
• Charge 47.87 million iPhones daily for an entire year
  
  Also read: Killer or a breeder? AI's impact on creativity and jobs
  

Training Costs of ChatGPT

Training large-language models (LLMs) is a highly energy-intensive process. During this phase, the AI model ‘learns’ by analysing large amounts of data and examples. 

• Training OpenAI’s GPT-3 with 175 billion parameters took about 34 days and used 1,287,000 kilowatt-hours
• In contrast, training GPT-4 with over 1 trillion parameters consumed around 62,318,800 kilowatt-hours over 100 days, 48 times more than GPT-3

Source: AI’s Power Demand report by BestBrokers.com