Chat GPT-4 Consumes Three Bottles of Water per 100 Words

OpenAI’s GPT-4 model, known for its cutting-edge AI capabilities, is under scrutiny for its environmental impact. A recent study by the University of California, featured in The Washington Post, reveals that generating just 100 words with GPT-4 consumes roughly the equivalent of three bottles of water.

This significant water usage has raised concerns about the sustainability of AI as it continues to expand, requiring more resources. The study also found that water consumption for cooling data centers varies by location. For instance, Texas uses 235 milliliters of water per 100 words, while Washington reported the highest usage at 1,408 milliliters.

This high consumption is linked to the cooling systems necessary for data centers that power AI models like GPT-4 and Copilot. The growing power needs of AI systems developed by companies like Microsoft and OpenAI are also putting pressure on local water resources and energy demands.

Reports indicate that OpenAI’s operational costs, including cooling, are immense, with estimates suggesting that running ChatGPT costs up to $700,000 per day. Despite significant investments from major tech companies, OpenAI faces financial hurdles, with potential losses projected to reach $5 billion within the next year.

Additionally, there are concerns over the availability of AI chips, particularly from NVIDIA, as demand surpasses supply. In response, Microsoft and OpenAI are reportedly investing $100 billion in the “Stargate” project to reduce their reliance on NVIDIA chips.

Elon Musk has also warned that while AI represents a groundbreaking technological advancement, there may be a global power shortage by 2025, unable to sustain its rapid growth.

As the AI industry expands, concerns about sustainability are becoming more urgent. Microsoft’s collaboration with BlackRock to raise $100 billion for data center expansion aims to address power constraints that could hinder AI development. However, finding environmentally sustainable solutions remains a critical challenge for the sector.