Generative artificial intelligence[i], represented by tools such as ChatGPT, Copilot, or DeepSeek, are rapidly becoming part of everyday life. Whether used in professional or personal contexts, this technology feels unstoppable. Yet behind the excitement lies a crucial question: what is its real climate cost? Is AI the main driver of digital emissions, or are other elements, like video streaming, more impactful?

To assess the climate impact of AI, two stages must be distinguished. The first is model training, an initial phase that consumes vast amounts of energy. For example, some reported advanced AI Models, required about 1,287 MWh of electricity, equivalent to the annual consumption of 120 U.S. households; and generated nearly 552 tons of CO₂e, comparable to the yearly emissions of more than 110 cars. Millions of liters of water were also used to cool servers during this process[ii].

The second stage is daily use by millions of users. Each query consumes on average 0.3 to 3 Wh of electricity and emits 2.5 to 5 g of CO₂e[iii], which is equivalent to burning a sheet of paper. While the impact of a single request seems minor, the cumulative effect of billions of queries is far from negligible. Estimates suggest that between twenty and fifty queries require about half a liter of water for cooling data centers.

Placed in the broader context of digital technologies, these figures are striking. In 2024, the information and communication technology (ICT) sector accounted for about 2 to 4% of global greenhouse gas emissions and 6 to 7% of worldwide electricity consumption, a share comparable to the aviation industry[iv]. According to the International Energy Agency, electricity use by data centers could double by 2030, driven largely by the growth of AI[v].

Streaming video, however, remains the largest source of internet energy consumption today. Between 60 and 70% of global traffic is linked to platforms such as YouTube, Netflix, TikTok, and other video services[vi]. Based on studies, one hour of HD streaming across platforms can produce between 50 – 100 g of CO₂e. Social media platforms, with their short‑form videos, account for roughly 15% of web traffic and nearly 50 million tons of CO₂e annually[vii].

In 2025, online video is projected to account for more than 80% of global internet traffic[viii]. Yet projections show that generative AI will grow at a much faster pace by 2030. While streaming will continue to expand, its growth is expected to be slower than that of AI. Even so, video is likely to remain the largest share of web traffic by 2030.

Generative AI represents a major technological breakthrough, but its role in climate change cannot be ignored. Between the massive energy required for training and the cumulative impact of daily use, its effect is already comparable to entire industries. Innovation must therefore be rethought considering sustainability, ensuring that progress does not come at the expense of the planet.

The impact of AI and video streaming on climate change is not only the responsibility of the companies that develop and operate these technologies, but also of the millions of users whose collective actions create significant effects. If we truly want to save our planet, we must ask ourselves whether we are ready to adjust our daily habits to reduce the footprint of our choices or at least rationalize them. As the saying goes, small stre