The hardest part is no longer getting the numbers. It's deciding what to do with them. The post Our AI Has an Invisible Footprint — Let's see How to Measure It appeared first on Towards Data Science. Some years ago, when I first joined the world of AI, I was super interested in measuring the environmental impact of the projects I was developing in my team. I recall being a regular user of Code Carbon, an open-source French library which can be easily embedded into your code to measure the energy consumption and carbon emission of your machine learning training. You start a tracker, you run your training loop, you stop the tracker, and you get the kWh and CO2e emission of your experiment. I had embedded it into all my projects. Measuring not only the performance but also the emissions of models felt like a concrete step we could take to make a difference. Two weeks ago, discussing some aspects of this tool with a colleague, I realized with some embarrassment I could not give an answer to some of their questions. And somehow as agents and generative AI have been a growing part of our daily business, their environmental footprint has been left behind — not because of a lack of tools, but because the role of our work as AI/ML engineers has evolved considerably. And that motivated me to reflect about the possibilities we have of bringing this good practice back into our workflows, so that the projects we work on in 2027 take this dimension into account. If you started working in data science a few years ago, like me, you probably remember a time when we built and ran our own models. We chose the data, trained the models, and deployed them on hardware we could see and understand. In 2026, many of us have become users of AI systems rather than builders of the models themselves. We call an API, chain prompts, connect an agent to tools, or switch on an AI feature built by someone else. The computation happens on machines we never see, in data centres we cannot name, running models whose size is often a secret. In this article, I want to take you on a journey with me to explore how we can step back from our projects, break them down into smaller components, and find ways to evaluate their environmental impact. Not perfectly, but well enough to establish where we can make a difference. To make this concrete, I'll follow one common use case throughout the article: a RAG assistant built on a hosted model.
1. What do we mean by environmental impact? When discussing the footprint of artificial intelligence (AI), the conversation typically centers on its carbon footprint. However, carbon is only one of five major impacts, which are as follows:
- Energy: the electricity consumed by machines, expressed in watt-hours (Wh) or kilowatt-hours (kWh).
- Greenhouse gas emissions: expressed in grams or kilograms of CO2 equivalent (CO2e). These depend on how the electricity is produced: for a given amount of energy (kWh), emissions are significantly higher if the electricity comes from coal rather than from nuclear or wind power.
- Water: used for cooling data centers (via evaporation) as well as for electricity generation.
- Materials: metals and minerals extracted for the manufacturing of chips, servers, and network equipment.
- Electronic waste: the fate of IT hardware once it is replaced.
These impacts occur at various stages of a machine's life cycle—from manufacturing, through its period of use, to its end-of-life. A comprehensive analysis of these stages and impacts is known as a "life cycle assessment" (LCA).
In this article, we focus exclusively on the usage phase—that is, the electricity consumed by machines while running AI, along with the resulting carbon emissions and water consumption. We do not account for hardware manufacturing or end-of-life disposal. This is not because these aspects are unimportant, but because they are outside the scope of this discussion.
Source: Towards Data Science · Summarized by HeadlinesBriefing