We need to talk about AI and Sustainability. Seriously.
Everyone is talking about how they use AI to become more efficient, proficient, how companies may (or may not) layoff workforce because of AI, their prompts that enable them to do this and that, the newest shiniest model from big-tech vendor X, Y or Z and so forth.
Monopolising Human Knowledge
AI companies train their models on all public works ever done by humans, available over the internet – even some copyrighted materials, material behind paywalls and more. This is a problem. 117 billion people have lived, learned and created new knowledge or art based on their life experience. Their life is the legacy of all of humanity.
A few big-tech companies are working on monopolising the access to this Legacy of Humanity. They do this by making their models available to most of us, and through convenience. Bernie Sanders have suggested those companies to be owned 50% by a publicly owned sovereign wealth fund. A brave idea, that probably won’t stand a chance.
Summary
5-7 minutes to read
The current rapid expansion of AI risks a bleak future: a handful of corporations could monopolise truth, control citizens, and dismantle democracy.
Meanwhile, the insatiable demand of data centers for water and electricity, may push these resources out of reach for ordinary people, leaving societies passive and disempowered.
If we act now, by engaging in public dialogue, pressuring our elected representatives, and demanding sustainable practices, we can ensure AI serves as a force for good, benefiting everyone.
What future will you choose?
Privatising The Truth
The AI companies want to become the gateway to human knowledge. The library of the future. And they will control the access to our common knowledge.
But perhaps the models will also slightly modify the knowledge, especially over time. And perhaps they will hide knowledge from some users, either by design or accident. The opacity and non-determinism of how the models work, allow the AI companies to be excused.
Over time, what used to be objective and immutable historical records, and even language itself may be changed in order to manipulate. Democracy will become an illusion, AI have turned our society into a totalitarian state. Congratulations, Larry Ellison. Welcome to 1984.
Mountains of Garbage
If this wasn’t bad enough, what used to be the New Rules for the New Economy – stating that the future will belong to services, not production resources – have proven wrong with AI. Data centers for AI are being built at an extreme scale. Meta’s Hyperion data center in Louisiana will be 2,250 acres – or more than 9 square kilometers. They have secured 1,400 adjacent acres more, totaling almost 15 square kilometers. And that’s just 1 datacenter, 1 company.
And then there’s Oracle, OpenAI, xAI (SpaceX), Microsoft, Google, Amazon..
A recent report from the UN University projects, that the yearly amount of AI-related electronic waste will be 2.5 million tonnes in 2030 – comparable to 250 Eiffel towers.
Part of the reason being the ultra fast development cycle of the GPU’s driving the AI workloads. Today’s fastest, AI GPU, will be replaced by its successor in 6 to 12 months. Just one xAI datacenter, today containing 200,000 GPUs, is planned to be expanded to 1 million GPUs.
Can this waste be efficiently recycled? Probably not. Perhaps someone will buy it? Otherwise, where do we put it? The externalities of the AI hype is extreme.
Oceans of Water
Another big problem of these data centers, is the amount of fresh water they consume for cooling. They deplete the same water sources, as people use for drinking. The xAI data center in Memphis was promised to recycle water. Now it doesn’t. It draws several million gallons of water per day – more than 9 million litres, or 9000 cubic meters. Again, we’re only talking about a single data center.
In Chile the population have tried to stop new data center developments to no avail. What used to be wetlands resembles desserts., according to The Guardian.
At the same time, Chileans was promised lots of new jobs related to the data centers. Only a fraction of those jobs became real.
Gigawatts of Electricity
In many countries, the electric grid can’t cope with the AI data centers buildup. In 2024 in Ireland 22 percent of the national consumption of electricity was appointed to data centers. It’s probably more today. From 2015 to 2023 this has caused electrical bills in Ireland to increase by 715 million Euro, which 11 percent of the population haven’t been able to pay.
We have truly entered the trap of Success to the Successful. Those who have money don’t care, they continue expanding their capacity, while they get richer and everyone else gets poorer.
A Positive AI Future
We can’t rely on a single vendor, or on AI dominated by the USA or China. AI is a tool we need to solve old and new problems, to help humanity thrive. We’re not Luddites; we embrace progress, but we demand it serve everyone.
To build AI differently, we must:
- Prioritise sustainability: Develop cooling methods that recycle cooling water, or risk turning our land into deserts, where food and life cannot flourish.
- Empower local communities: Support AI that runs locally. Companies like Mistral and others now offer open-weight models, that can be downloaded and run on local hardware. This decentralises truth, reduces costs, and conserves land, water and electricity, ensuring equitable access to resources.
- Demand true openness: Most open models are open-weight, not open-source. The Allen Institute for AI (AI2) is a rare exception. We need more vendors to follow their lead.
The European Tech Sovereignty Package addresses some of these goals, but its success depends on us.
Let’s work together toward a future where AI is green, democratic and accessible, where the Legacy of Human Knowledge belongs to all of us, not just a few big companies.
If we act now, we can make this future a reality!
This blog post was inspired by the work of many great people, most lately an article from the Danish newspaper, Information.
Previously, we also read an article from The Vatican, signed by Pope Leo, that AI must serve humanity, not concentrate power.
Our previous blog post on Digitally Sovereign AI (February 2026).
Dette blog-indlæg er skrevet på engelsk, da det er et fælles-europæisk problem, vi gerne vil have så mange som muligt til at fokusere på. Nyd folkemødet, jer der har tid til at tage derned 😊
Skriv et svar