environmental impact of AI
Welcome! It's great to have you here. Before we dive into the data-heavy aspects of how artificial intelligence impacts our planet, I'd love to learn a bit more about you so I can make this session as relevant and engaging as possible.
To help me tailor our session, could you tell me a bit about your background? For example, are you coming at this from a technical perspective, an environmental studies interest, or just general curiosity about the topic?
I'm just generally curious about how AI affects the planet
It's wonderful to have you here! Since you're curious about how AI affects our planet, we've designed a clear path to explore this together. We'll start with a bird's-eye view of where these environmental impacts come from, look closely at the massive energy and water demands, and wrap up with how we might use AI itself to build a greener future.
Your path
Deep Dive: Energy and Carbon Emissions
Hidden Impacts: Water, Land, and Minerals
The Double-Edged Sword: AI for Sustainability
Future Outlook and Mitigation Strategies
The Lifecycle of an AI Model
To understand how AI affects the planet, we have to look at its entire lifespan. We often think of AI as living entirely in a weightless, virtual cloud, but every search query, generated image, and translation relies on massive, physical, power-hungry machinery.
We can break down this physical footprint into four key stages of an :
| Lifecycle Stage | What Happens | Environmental Toll |
|---|---|---|
| Design & Mining | Extracting raw materials like silicon, lithium, and cobalt. | Severe habitat disruption, mining pollution, and carbon-heavy manufacturing. |
| Training | Feeding a brand-new model billions of words or pictures so it can learn patterns. | Months of continuous supercomputing, using enormous amounts of electricity. |
| Deployment & Usage | Running the model in data centers so millions of people can query it daily. | Constant power draw and water cooling for thousands of active servers. |
| Disposal | Retiring outdated server hardware after a few years of intense wear and tear. | Rapid accumulation of toxic electronic waste (e-waste). |
While tech companies often focus on the energy it takes to train a model, everyday usage—which scientists call inference—actually makes up the vast majority of AI's ongoing power draw. Every single prompt we type has a tiny, cumulative cost.
When we look at the physical impact across this lifecycle, it generally boils down to three main categories: Carbon (from electricity use), Water (used to cool hot server rooms), and Waste (from discarded microchips and hardware).