The Hidden Cost Behind Every “Thanks”
In our everyday lives, saying “please” and “thank you” is a basic courtesy — a sign of politeness we take for granted. But what if we told you that those tiny, kind words are costing us millions of dollars, gigawatt-hours of electricity, and even millions of liters of fresh water?
Welcome to the curious case of how polite language — when typed into AI systems — has a measurable environmental footprint.
According to OpenAI CEO Sam Altman, people’s habit of politeness with AI is surprisingly expensive and energy-intensive. While charming, these extra tokens (words) place a real-world burden on energy grids and water resources.
Let’s dive into how something as small as a “you’re welcome” can add up to a massive sustainability problem and what we can do about it.
How Polite Language Affects AI Systems
Every word you type into AI language models is processed by large-scale computational systems. These systems — hosted in data centers — consume enormous amounts of electricity to run powerful GPUs and even more resources to keep those systems cool.
Here’s the key: “Please,” “thank you,” and similar phrases may seem tiny in your prompt, but on a global scale, their impact is massive.
Sam Altman recently shared that processing these polite interactions can cost tens of millions of dollars every year. Each polite prompt adds a few tokens, resulting in more computation, increased electricity usage, and greater strain on infrastructure.
Let’s quantify this.
Environmental Statistics: The Real-World Impact of Being Polite
Global Usage Snapshot
- Estimated Monthly Active AI Users: 30 million+
- Approx. 70% of users regularly say “please” and “thank you” in prompts
- That’s 21 million users engaging in polite prompts
Prompt Frequency & Annual Polite Prompts
- Avg. interactions/user/day: 10
- Daily polite prompts: 21 million × 10 = 210 million
- Yearly polite prompts: 210 million × 365 = 76.65 billion
Electricity Consumption
- Avg. energy per prompt: ~0.03 kWh
- Extra energy for polite language (~5–10% more tokens): ~0.004 kWh
- Total extra energy/year: 76.65B × 0.004 kWh = 306.6 GWh
To put that into perspective:
- That’s enough to power 28,000 U.S. homes for an entire year.
- Equal to running 50,000 refrigerators continuously.
Water Usage for Cooling
- Avg. water needed for cooling per kWh: ~1 liter
- Total water consumption/year: 306.6 million liters
That’s the equivalent of:
- 122 Olympic-sized swimming pools.
- 3+ million cups of coffee (in cooling cost alone).
CO₂ Emissions
- Global average CO₂ per kWh: ~0.4 kg
- Total CO₂ from polite prompts: 306.6M kWh × 0.4 kg = 122,640 metric tons.
That’s equivalent to:
- Emissions from 27,000 gas-powered cars annually.
Financial Impact
- Average cost per polite prompt: ~$0.0004.
- Total cost annually: 76.65B × 0.0004 = $30.6 million.
Why Does Politeness Cost So Much?
The expenses associated with polite prompts can be attributed to various processing steps:
- Breaking sentences into tokens (each word = ~1.3 tokens).
- Processing these tokens through multiple neural layers.
- Storing and returning the output, which often includes another polite response.
For example, when you write:
User: “Can you explain that again, please?”
AI: “Of course! Here’s a simpler version…”
User: “Thanks.”
AI: “You’re welcome!”
You’re actually initiating two extra prompts and multiple processing cycles for manners. Multiply this by billions, and it becomes a major environmental concern.
What Is Green Computing?
Green computing aims to reduce the ecological impact of digital technologies. It emphasizes:
- Energy-efficient hardware.
- Renewable energy use in data centers.
- Efficient algorithms and code optimization.
- Water-saving cooling systems.
According to various studies, green computing is essential for aligning IT development with environmental goals, especially in power-hungry fields like artificial intelligence.
How Can AI Become More Sustainable?
- Smart Politeness Detection: Future models could detect polite intent without parsing every word by using sentiment analysis to infer manners.
- Templated Responses: Predefined token-light responses, such as emojis or brief phrases, could lower energy usage.
- Infrastructure Optimization: AI labs can offset polite costs by:
- Running servers on renewable energy sources.
- Implementing advanced cooling systems.
- Automating resource management when not in use.
Should You Stop Saying Thank You?
Not necessarily. It’s not about halting kindness — it’s about being aware of the digital footprint.
We don’t need to stop being polite; we need to design systems that recognize politeness more efficiently. In the meantime, users can be mindful of their word count in interactions.
Examples of Mindful Communication
Instead of:
“Could you please explain this again in a bit more detail if possible? Thanks in advance.”
Try:
“Explain again with more detail.”
By doing so, you’ll receive the same output but with reduced resource waste.
Final Thoughts: Mindfulness in the Age of AI
Digital behavior might feel intangible, but it has very real consequences for the planet. While kindness in human-AI interactions is admirable, we’re uncovering that even digital manners have their costs.
The positive takeaway? Through smarter design and energy-conscious usage, we can work toward making AI more sustainable without sacrificing its human touch.
So go ahead and say “thank you,” just know: even digital politeness can impact our planet.
