AI has a real environmental footprint. You get to choose how big yours is.

Is AI bad for the environment? The honest answer

AI has a real footprint. The models run in data centers full of specialized chips, those chips draw serious electricity, and the servers need cooling, which often consumes water. Training a large model is a one-time energy bill the size of a small town's, and serving billions of requests every day adds up to far more over time.

But "AI" is not one thing. The footprint of a request depends on the size of the model, the efficiency of the data center, and above all on the energy powering it. The same question answered by a right-sized open source model in a hydro-powered Swiss data center and by a giant model on a gas-heavy grid are two very different events for the planet. The problem is not AI itself. It is how and where it runs.

How much energy and water does AI actually use?

The honest numbers are estimates, because most providers do not publish them, but the orders of magnitude are well documented:

• Energy per requestA single AI answer costs roughly ten times more electricity than a classic web search. Small on its own, enormous multiplied by billions of daily requests.
• Water for coolingData centers evaporate water to stay cool. Researchers estimate that a session of a few dozen AI questions can consume the equivalent of a small bottle of fresh water, depending on the site and the season.
• Training costTraining a frontier model consumes gigawatt-hours of electricity and millions of liters of water. This bill is paid once, then amortized over every request the model serves.
• The trendData centers already account for a meaningful share of global electricity demand, and AI is the fastest-growing part of it. That is exactly why the energy source behind them matters.

Green AI: the renewable compute answer

A new category of providers has decided that AI should run on clean energy, and some go further than just buying green power. A few real examples working today:

• Euria (Infomaniak)A free, sovereign AI assistant from Swiss cloud provider Infomaniak. It runs open source models in data centers powered by renewable energy, and the heat from the servers is recovered to warm homes through Geneva’s district heating network. The same energy works twice.
• GreenPTA privacy-friendly AI service focused on sustainable inference, making the hidden energy cost of each AI request visible instead of ignoring it.
• CrusoeAn energy-first AI infrastructure company that builds data centers where clean or stranded energy already exists, instead of adding load to dirty grids.

Why open source models fit green compute

Green providers almost always run open source models, and that is not a coincidence. Open source models come in many sizes, so a provider can pick one that is right-sized for the task instead of pointing a giant model at every question. A smaller model that answers well uses a fraction of the energy and water.

Open source also means the model can run anywhere: a Swiss data center heated by hydropower, a wind-powered site, or your own machine. When the model is a portable artifact rather than a locked API, whoever runs it gets to choose the energy behind it. That choice is exactly what proprietary AI takes away.

How this connects to owning your AI

AskMojo's core belief is that your AI setup should be yours: portable skills, exportable knowledge, no lock-in to a single provider. Energy is part of that same freedom.

If your skills and knowledge base are plain, portable files, you can point them at whatever engine you want, including one powered by renewable energy. The green AI movement and the own-your-AI movement are two sides of the same idea: you decide how your AI runs, where it runs, and what it costs the world around you.