MIT Explained: Generative AI's Environmental Impact

After reading the MIT News article on the environmental impact of generative AI, I found myself thinking a lot about the hidden side of “the cloud.” I usually picture AI as this clean, digital thing, type a question, get an answer, but the article is a reminder that there’s a very real physical cost behind it. Massive data centers, huge energy demand, tons of water for cooling, all of that is part of the story every time generative AI is used. And I completely agree with the author on this point, AI isn’t magical. It runs on electricity, and right now, that energy usually comes with a carbon footprint attached.

I also appreciated the way the article pushed back against the idea that only training these big models is the problem. The environmental impact continues every single time someone runs a prompt. Inference, the part we interact with, is happening constantly and everywhere. It’s easy to forget that because the burden is invisible to the user. I think the article makes that point really well, and it’s something most people probably don’t realize.

That said, there were a few things I wish the article had gone deeper on. For example, the piece focuses heavily on the costs of generative AI, but it barely touches on the benefits. AI isn’t just about writing essays, it’s also being used to optimize supply chains and model climate scenarios, and improve efficiency in ways that could reduce emissions. I’m not saying that cancels out the environmental footprint, but I do think a fair discussion needs to look at the full picture, not just the downside.

Another point where I felt the article oversimplified things is around energy sources. Yes, data centers use a lot of power. But not all data centers are equal. Some run on mostly renewable energy, some use advanced cooling systems, and the efficiency of AI hardware is improving really quickly. The article acknowledges this a little, but it mostly sticks to a “things are getting worse” narrative. In reality, there’s a lot of innovation happening right now.

I also wish the article had talked more about accountability. Who should be responsible for managing AI’s environmental footprint? The companies building these models? Governments? Users? Without clear reporting standards for energy use, water consumption, and emissions, it’s almost impossible to know whether we’re actually making progress. Transparency is a huge missing piece in this conversation.

But overall, the article made me think, in a good way. It highlights a problem that’s easy to ignore because it feels distant and abstract. It’s a reminder that digital tools still have real-world consequences. At the same time, I’m hopeful. If we push for better standards, better reporting, and better technology, AI doesn’t have to be at odds with environmental goals. It can actually help us reach them. The challenge is making sure we’re honest about the trade-offs, and willing to design systems that don’t hide the cost.

Reference: Explained: Generative AI’s Environmental Impact,  https://news.mit.edu/2025/explained-generative-ai-environmental-impact-0117