Fashion's Next Hidden Cost: AI and the water nobody's counting

Is artificial intelligence really making fashion more sustainable, or are we moving too fast to ask the right questions?

I am writing this on a laptop, using AI to help me come up with a title, while writing about the true cost of the very tool I am using. The irony is not lost on me.

A few weeks ago, I found myself deep in research on how we can incorporate AI technology into the handicraft and handloom sector of India. I know the gaps between artisans and technology well, the distance is vast, and the bridge is rarely built with the artisan in mind. I was thinking about new and accessible ways to make AI useful for craftspeople and their next generation, so that they too have something to look forward to in terms of learning, developing skills, and keeping their crafts alive.

What I found was genuinely exciting. Ideas at trial stages across India, ideas at the ideation stage, and some that seemed worth pursuing seriously. AI is already reshaping apparel forecasting, manufacturing, and waste reduction in ways that are hard to ignore. Introducing it to the craftspeople, so that their process could become more efficient and the future of their work more secure, seemed like the natural, responsible next step.

"And then, in the middle of my research, I found something I was not looking for."

To keep a data center running, fresh water is used to cool the systems that power every AI tool, every query, every response. The scale of this is not something that is widely discussed, and when I encountered it for the first time, it stopped me entirely.

The numbers, from primary sources:

Google - Consumed 8.1 billion gallons of water across its data centres in 2024, an 88% increase since 2019. A single data center in Iowa used 1 billion gallons in 2024 alone: enough to supply all of Iowa's residential water for five days. (Google 2025 Environmental Report)

Microsoft - Global operations consumed approximately 1.69 billion gallons of water in its most recent reporting year, a 34% increase from the year before. (Microsoft Environmental Sustainability Report)

Meta - Consumed 813 million gallons of water globally in 2023, with 95% used by data centers. (Meta 2024 Sustainability Report)

India - Data centers in India consume 26 million liters per MW annually, resulting in nearly 1.4 crore liters of daily water usage in Bengaluru alone, equivalent to the daily water needs of approximately 41,900 households.

Note: These figures cover direct water use for cooling only. The indirect water required to produce the electricity that powers these centers is estimated to be 12 times greater, and is rarely included in corporate disclosures.

The technology being proposed to make craft industries more efficient, more sustainable, more future-ready, runs on water that communities, including artisan communities, cannot afford to lose.

This is the irony that nobody is talking about. Or more precisely, not enough people know about it. The same technology framed as the answer to fashion's sustainability crisis carries a hidden environmental cost that is growing faster than our understanding of it. And then there is our artisan community, the most vulnerable people in this entire conversation, the ones least likely to be informed before a tool is placed in their hands and imparted onto their livelihood.

To be fair, the companies are responding. Google replenished 64% of its freshwater consumption in 2024 and has committed to replenishing 120% by 2030. Microsoft is piloting zero-water cooling systems, with new designs beginning to come online. These are not small commitments. But the gap between commitment and current reality remains wide, and the people making decisions about AI adoption in craft sectors are largely unaware that this gap exists.

The fashion industry is already on a path toward polluting less, wasting less, building more carefully. The incorporation of AI into that journey should be understood fully, not adopted because it is available, not promoted because it sounds progressive, and not handed to the most vulnerable parts of the ecosystem without explaining what it runs on.

Before we hand this technology to someone

whose livelihood, community and water supply

may all be connected,

Do we have a responsibility to explain the full picture?

I think we do. And I think that responsibility begins with asking the question out loud, even when, especially when, we are already using the tool ourselves.