AI models may seem like immaterial digital tools—but they come with a very real environmental footprint. From the massive energy required to train large language models (LLMs) to the ongoing cost of everyday queries, understanding the ecological impact of AI is key to responsible use.

Training Large Models

Training a state-of-the-art model like GPT-3 reportedly consumed around 1,287 megawatt-hours of electricity and produced 552 metric tons of CO₂ emissions—comparable to the yearly emissions of 123 cars (Strubell et al., 2019). More recent models like BLOOM have aimed to improve efficiency, but training remains energy-intensive.

Daily Usage and Inference

Every query you run also uses energy, though much less than training. A single ChatGPT query consumes about 0.001–0.003 kWh, which translates to 0.1–1.3 grams of CO₂ depending on the model and infrastructure (Devera.ai, 2024). That may seem small, but multiplied by billions of daily queries, the footprint grows rapidly.

Why It Matters

The environmental cost of AI isn’t just about carbon emissions. It also involves water use for cooling data centers and rare materials for hardware. As demand for GenAI rises, these hidden costs must be considered alongside its benefits. Awareness helps users and developers alike make choices that balance innovation with sustainability.

AI has enormous potential—but it also consumes enormous resources. By recognizing its ecological cost, we can push for more efficient models, greener energy sources, and mindful usage. Responsible AI is not just about fairness in language, it’s about sustainability for the planet.

We’ve looked at the cost to the environment—next, let’s talk about the cost to users.
In the next article, we’ll explore the financial cost of AI: how tokens translate into money, and why it matters for accessibility.

👉 Read next: The Financial Cost of Using AI