Angelina YangFounder and Host, TwoSetAI · AI builder for visibility and discoverability

AI's Hidden Cost: 5 Million Liters of Water to Train GPT-3

30 Apr 2026 · 0 min

Watch on YouTube

0:07

In this brief segment, Angelina Yang highlights a striking environmental claim that training OpenAI's GPT-3 consumed over 5 million liters of fresh water. She explains that traditional data‑center cooling systems rely on large volumes of water, which is pumped through the facility, absorbs heat, and then evaporates, leading to substantial water consumption during AI model training and deployment.

OpenAI, when they're training GPT-3, they're using over 5 million liters of fresh water for cooling the data center.
Cooling in data centers, it requires, at least the more traditional ones, require a lot of fresh water to be pumped into the data centers, which is then, you know, which absorbs the heat and is evaporated out. So you consume a lot of water in cooling the data centers where AI models are being trained or deployed.