Why this exists

I work in AI. People ask me about AI's water usage all the time. I see posts on Instagram and Twitter with alarming numbers but no context. No sources, no comparisons, no way to tell what's real. So I went looking for the truth. When I started digging into what AI companies actually disclose about water usage, I found two things.

First, the information is scattered and hard to verify. Some providers publish detailed sustainability reports. Others publish nothing. Finding out which is which means reading hundreds of pages of corporate reports, and even then the picture is incomplete. Meanwhile, the conversation swings between "AI will drain our aquifers" and "it's fine, don't worry." Neither is grounded in what's actually published.

Second, no one had assembled the disclosure data into a single, maintained, citable reference. Carbon tracking has tools, dashboards, and standards. Water transparency has scattered sustainability reports and paywalled research.

AI can be a net benefit to the world. But that case is stronger when the industry is open about its resource consumption. Transparency isn't an attack. It's the foundation for trust. And right now, the gap between what providers know and what they publish is wide.

This project addresses that gap. Not by calculating water usage or assigning scores, but by documenting what each provider actually discloses, with a source link for every claim.

What this is

A public record of environmental transparency across the AI industry. Not a score. Not a ranking. Just a factual accounting of what each provider chooses to publish and what they don't. The dataset is open source, schema-validated, and licensed under CC BY 4.0 so researchers, journalists, and developers can use it freely.

What this is not

This is not an environmental impact assessment. We do not estimate total water usage for providers who don't report it. We do not assign scores or grades. We report what is disclosed and cite the source.

How it's built

This site was built with open source code, ships zero JavaScript, and produces static HTML. No servers, no tracking, no cookies. The data is maintained as a JSON file validated against a schema on every change.

It was built using Claude Code (Anthropic's Opus 4.6 model). We disclose this for the same reason we expect AI providers to disclose their water usage: transparency is the point. Full details in our transparency disclosure.

This is a starting point

Every claim is sourced from published sustainability reports, peer-reviewed research, and official disclosures. But this is the beginning of something we believe should exist, and it's better built in the open where others can verify, improve, and expand it.

If you spot an error, open an issue. If a provider has published data we missed, submit a PR. If you're a researcher with better sources, we want to hear from you.