The Real Cost of AI: Does Politeness Really Waste Energy?
A curious notion has been circulating online: that using polite words like “please” and “thank you” when interacting with ChatGPT consumes extra electricity and harms the environment. The argument sounds logical at first glance—more words mean more processing, and more processing means more energy. Even Sam Altman, the head of OpenAI, has acknowledged that at the scale of billions of queries, these costs add up. Yet the reality is far more nuanced. The energy consumed by a handful of extra words is negligible compared to the massive data centers that power AI systems like ChatGPT. Still, this idea points to something important: people are beginning to recognize that artificial intelligence is not magic. It runs on real machines, real electricity, and real resources—and those have a tangible impact on the world.
Data Centers and Infrastructure: The True Energy Guzzlers
Artificial intelligence relies on sprawling data centers packed with high-performance computers. These machines consume enormous amounts of electricity. Keeping them cool requires constant water and energy for cooling systems. That is why AI is directly tied to resources like power, water, and land. As AI usage grows, so does its environmental footprint. The real issue, then, is not whether we type polite words into a chatbot. It is how often and how intensively we use these systems in the first place.
Why Every AI Query Has an Energy Cost
There is a crucial difference between AI and everyday digital services like video streaming. Traditional software mostly retrieves existing data—opening a document or watching a video involves minimal new processing. AI models work differently. When you ask a question, the system generates a fresh response each time. This process, called “inference,” requires the AI to perform new calculations for every query. Each request consumes energy anew. That is why AI behaves more like a large infrastructure than standard software. The more we use it, the more energy it demands.
Beyond Electricity: Water and Other Resources
The International Energy Agency has warned that data center electricity consumption could double by the end of this decade. But electricity is only part of the story.
- Water: Massive amounts of water are needed to cool the servers inside data centers.
- Land and materials: Building and operating these facilities requires significant land and various raw materials.
Although AI services are used globally, their impact is most pronounced locally. Data centers place additional strain on the electricity and water supplies of the communities where they are located.
Why This Myth Matters
Consider countries like New Zealand, where most electricity comes from renewable sources. Even there, new data centers put pressure on the local power grid. The reason is simple: electricity used to run servers is electricity that cannot be used for anything else. So instead of viewing AI as just a digital convenience, we should see it as a physical system that demands real power and resources.
The idea of skipping “please” and “thank you” when talking to ChatGPT may not have a significant scientific impact on the environment. But it is a meaningful signal. It shows that people are waking up to the fact that AI has an environmental cost. Taking this awareness seriously is essential for deciding how to integrate AI infrastructure with our energy plans and society at large. There is no need to reject AI—it is proving immensely valuable in fields like research and healthcare. But like any industry, it comes with both benefits and drawbacks. If we treat AI as nothing more than invisible software, we will fail to understand its true cost and real-world consequences.
