[nagonago]

The problem is not so much the carbon footprint of just using LLMs, but the impact from building and running massive data centers needed to train the models. Check out this report from the ITU:

https://www.itu.int/en/mediacentre/Pages/PR-2025-06-05-green...

> According to the latest edition of the report, electricity consumption by data centers — which power AI development and deployment, among other uses — increased by 12 per cent each year from 2017 to 2023, four times faster than global electricity growth.

> Four leading AI-focused companies alone saw their operational emissions increase in the reporting period by 150 per cent on average since 2020. This rise in energy that is either produced or purchased – known as Scope 1 and Scope 2 emissions – underscores the urgent need to manage AI's environmental impact.

> In total, the amount of greenhouse gas emissions reported by the 166 digital companies covered by the report contributed 0.8 per cent of all global energy-related emissions in 2023.

It's worth noting that it's not all gloom and doom. As the report optimistically notes:

> Renewable energy adoption: 23 companies operated on 100 per cent renewable energy in 2023, up from 16 in 2022.

> Dedicated climate reporting: 49 companies released standalone climate reports, signaling greater transparency.

> Scope 3 consideration: The number of companies publishing targets on indirect emissions from supply chains and product use rose from 73 to 110, showing increasing awareness of industry impacts.

So the environmental impact is alarming, but at least companies seem to be growing more conscientious about it. We must continue to hold tech companies accountable for their environmental impact, and keep pushing for renewable energy.