In a recent LinkedIn post, Ruben Hassid challenges the narrative that Artificial Intelligence is a significant drain on global water resources. Hassid argues that the water footprint of AI is often exaggerated when compared to everyday consumer choices and agricultural practices.
Hassid begins by directly addressing the concerns about AI’s water consumption, stating:
“If you think AI is wasting water, you must stop: → Eating avocados (5x more water than every AI) → Drinking coffee (8x more) → Eating almonds (11x more) → Drinking beer, wine, spirits (16x more) → Cooking with olive oil (26x more) → Eating eggs (44x more) → Buying clothes (80x more) → Eating beef (84x more) → Eating cheese (170x more) → Eating bread (412x more)”
He further contextualizes AI’s water usage by comparing it to humanity’s total consumption. According to Hassid, all AI systems globally collectively use approximately 0.5 km³ of water per year, which he calculates to be about 0.05% of humanity’s total annual water usage.
Deconstructing the AI Water Footprint
Hassid delves into the technical aspects of data center cooling, a primary area where AI operations interact with water resources. He points out that older data center models often rely on evaporative cooling, where water is turned into steam and lost. He quantifies this loss:
“Old data centers cool their chips by evaporating water. It turns to steam, floats off, gone. Around 2.6 million gallons per megawatt, every year. Fresh water wasted because it’s used once.”
However, Hassid quickly pivots to highlight advancements in cooling technology. He explains that newer systems, like those from NVIDIA, utilize closed-loop liquid cooling systems. In these systems, a sealed loop circulates liquid, which is cooled by ambient air, significantly reducing or eliminating water evaporation.
The Shift Towards Water-Efficient Cooling
As Hassid notes, these modern cooling methods drastically cut down on water waste. He elaborates on the efficiency gains:
“The newest ones don’t do that. NVIDIA’s latest servers run one sealed loop of liquid. You fill it once. It runs hot water, hotter than a hot tub, so the cool outside air handles the rest. The same liquid just loops around. The result: close to zero water. Up to 100% less than the old way.”
This technological evolution, in Hassid’s view, suggests that the water concerns associated with AI are becoming less relevant as the industry adopts more sustainable practices. He implies that focusing solely on the historical impact of older cooling methods overlooks the rapid progress being made.
Consumer Choices vs. Technological Advancements
The core of Hassid’s argument is a call for perspective. He contrasts the water required for common food items and goods with the water used by AI. By listing items like avocados, coffee, almonds, beef, cheese, and bread, and their significantly higher water footprints compared to AI, Hassid aims to reframe the discussion.
He concludes with a satirical suggestion:
“So keep drinking almond milk coffee, with a piece of sourdough bread, wearing jeans, tweeting on your iPhone about AI water usage ruining everything.”
This closing statement underscores his point that many common consumer activities have a far greater water impact than AI, especially as the technology and its infrastructure evolve towards greater efficiency. Hassid encourages a more nuanced understanding of water usage, urging readers to consider the broader context beyond the immediate focus on AI.
📝 About This Content
This article is based on insights shared by Ruben Hassid on LinkedIn.
📅 Originally posted on July 1, 2026 | View original post on LinkedIn →