OpenAI’s GPT-4 model consumes 3 water bottles to generate 100 words

OpenAI’s GPT-4 model, known for its impressive advancements in artificial intelligence since the launch of ChatGPT in 2022, is facing scrutiny over its environmental impact.

A recent study from the University of California, highlighted by The Washington Post, reveals that GPT-4 consumes about three water bottles’ worth of water to generate 100 words. This has raised concerns about the sustainability of AI technologies as they continue to grow in scale and power consumption.

US mulls barring Chinese technology in vehicles on American roads

The study shows that water usage for cooling data centres varies by location, with Texas consuming 235 millilitres of water per 100 words, while Washington reported the highest, using 1,408 millilitres. This water consumption is tied to the cooling systems required for data centres powering AI models like GPT-4, Copilot, and others.

Last year, Jez Corden reported that Microsoft’s Copilot and ChatGPT consume nearly one bottle of water per query due to the energy demands for cooling, said an article from Windows Central. The increasing thirst for power by AI systems like those developed by Microsoft and OpenAI is raising concerns about local water resources and power demands, as AI continues to grow.

OpenAI’s daily operational costs, including cooling requirements, are staggering. Reports from last year indicated that running ChatGPT costs the company up to $700,000 per day. Despite heavy investment from major tech players like Microsoft, NVIDIA, and Apple, OpenAI has faced financial challenges, with some projections suggesting the company could face $5 billion in losses within the next year. However, additional rounds of funding have kept the company afloat, pushing its market valuation over $150 billion.

Nintendo sues hugely popular Palworld

Meanwhile, there are also concerns about the supply of AI chips, particularly from NVIDIA, as demand far outstrips supply. Microsoft and OpenAI are reportedly investing $100 billion in a project called “Stargate” to reduce their dependence on NVIDIA’s chips.

At the same time, Microsoft has partnered with BlackRock to raise another $100 billion to expand its data centres, hoping to overcome the power limitations that threaten to stifle the AI revolution. Notably, Elon Musk has warned that while AI represents the biggest technological breakthrough in history, the world may run out of electricity to power its advances by 2025. This growing demand for resources, including water and power, continues to spark debate about the future sustainability of AI development.