Nvidia and the Ecological Cost of AI: 2024 Sustainability Report Reveals +87% Emissions in One Year
Condividi questo post
Nvidia has quietly published its 2024 sustainability report. Few have talked about it, yet the main figure is as striking as it is worrying:
Total emissions rose from 3.8 million tonnes of CO₂e to 7.15 million.
An 87% increase in just one year.
Such a rapid rise is more than just a sign of one company’s success. It is the direct consequence of the global boom in semiconductor production and in the infrastructure required to power the unprecedented growth of artificial intelligence.
The AI Boom Is Reshaping Global Environmental Pressure
Behind the acceleration of generative AI sits an entire industrial value chain with extremely high environmental intensity — and not only in terms of electricity used by data centers.
1. Semiconductors Require Gigawatt-Scale Energy
Chip manufacturing plants — the famous fabs:
consume energy on the scale of entire cities,
need huge volumes of ultra-pure water,
generate complex streams of industrial wastewater that are difficult to treat.
The result: environmental impacts that are often underestimated, but very real.
2. Intensive Water Use and Hazardous Waste
Chip production requires:
millions of litres of water per day,
emissions of per- and polyfluoroalkyl substances (PFAS),
hazardous chemical waste,
industrial gases that can be up to 35,000 times more climate-warming than CO₂.
3. Ever-Larger Semiconductor Megaclusters
Some examples that should make us pause:
In South Korea, a 10 GW semiconductor cluster is under construction that will consume one-seventh of the country’s electricity and half of Seoul’s water.
In the United States, supported by the CHIPS Act, new fabs are opening with significant impacts on local communities.
4. The Future Energy Demand of AI Data Centers
According to Sam Altman (OpenAI), by 2033 AI alone could require 250 GW of data center capacity.
That is roughly equivalent to the entire installed power capacity of India today — around 19% of current global electricity demand.
And to produce the chips needed for this infrastructure?
It is estimated that the world would need at least 10 new semiconductor fabs, just to keep up with the demand from players like OpenAI.
It’s Not Just About Data Centers: The Whole AI Value Chain Weighs on the Planet
It is important to clarify one point:
The ecological footprint of AI is not only about the electricity consumed by data centers, as the media often suggest.
The real impact lies in:
extraction of critical minerals,
production of highly energy-intensive hardware,
chemically complex industrial processes,
massive infrastructure for networks, cooling, logistics and manufacturing.
On top of this come social and human risks:
workers’ exposure to toxic substances,
labour disputes and industrial conflicts,
governance issues in the regions where production is concentrated.
The Key Question: What Is AI Giving Us in Return?
Faced with such high ecological costs, one question is not only legitimate but necessary:
Does the value created by AI really justify the global environmental impact of the sector?
This is not about denying the usefulness of artificial intelligence. It is about assessing the true cost–benefit balance for society and the planet.
Technological innovation cannot be separated from environmental responsibility. We need to understand how to make this ecosystem sustainable, transparent and aligned with climate goals.
AI is not inherently “sustainable” or “unsustainable”.
It depends on how it is designed, produced and powered.
Nvidia’s 2024 report sends a strong signal: innovation without responsibility generates hidden impacts that we can no longer ignore.
The Real Issue: What Future Do We Want to Build with AI?
The question is not whether AI will be part of the future.
The real question is:
What kind of future do we want to build with AI?
One where emissions double in a year while we look the other way?
Or one where efficiency, social justice and planetary boundaries set the rules of the game for AI and the entire semiconductor industry?
The answer will not come from one company report alone — but it must start from here:
taking the ecological cost of AI seriously, and demanding clear commitments, measurable targets and transparent reporting from the companies that are shaping its infrastructure.
Christian Sansoni
