Google has issued a new study outlining the environmental impacts of its Gemini AI system, suggesting it consumes only ‘five drops’ of water per text prompt. However, experts such as Shaolei Ren from the University of California argue that this claim overlooks substantial water usage from indirect sources, primarily from the cooling systems in data centers linked to energy production.

Ren states, “They’re just hiding the critical information. This really spreads the wrong message to the world,” underlining the necessity to consider indirect resource consumption.

The International Energy Agency indicates that a significant portion of water used by data centers comes from auxiliary sources, like cooling the power facilities. Concerns arise around whether Google’s metrics sufficiently address these indirect usages.

Additionally, training AI models comes at a cost, both financially and environmentally, but Google hasn’t released any evaluations on these impacts. The conversation continues as industry experts analyze the broader implications of AI technologies on environmental sustainability.