Ever feel a little weird being rude to a machine? You’re definitely not the ones. Most of us find it naturally jarring to just bark orders at a screen without a little bit of courtesy. We’ve all done it—typed out a quick prompt and then felt the need to add a "please" or a "thank you" at the end, just in case the robot uprising starts and we want to be on the "nice" list.
But here’s the kicker. That politeness isn't free.
👉 See also: Inside a Space Shuttle: Why It Was Way More Cramped Than You Think
OpenAI CEO Sam Altman recently dropped a bit of a bombshell on X (the platform formerly known as Twitter) that caught everyone off guard. It turns out that when you say Sam Altman thank you to ChatGPT, you’re actually adding to a massive electricity bill. Like, tens of millions of dollars massive.
The High Cost of Being Nice to AI
It sounds like a joke, right? How could a simple two-word phrase cause that much damage? To understand it, you’ve gotta look at how Large Language Models (LLMs) actually work. They don't just "understand" words; they process tokens. Every single character, space, and punctuation mark you send over to the server requires computational power to process.
When millions of users are all being polite at the same time, those extra tokens add up fast.
Altman confirmed this in response to a user who wondered about the specific drain on resources. He admitted that the cost of these polite interactions has reached "tens of millions of dollars." But interestingly, he didn't tell people to stop. He actually called it "money well spent."
Why the bill is so steep
- Compute Power: Every "thank you" forces the model to run an extra inference step.
- Energy Consumption: Data centers are notoriously thirsty for power and water for cooling.
- Scaling Effects: With over 800 million weekly active users, a tiny gesture becomes a massive overhead.
Honestly, it’s a weird paradox. We’re programmed to be social creatures, and we carry those social norms into our digital lives. Even if the AI doesn't have "feelings" to hurt, our brains still feel better when we treat it like a person. Altman seems to value that human connection, even if it’s technically "wasted" compute.
Sam Altman Thank You: The 2023 OpenAI Saga
While the "thank you" memes are funny, there was a time not too long ago when the phrase Sam Altman thank you carried a much heavier weight. Back in November 2023, the tech world basically melted down for five days. If you weren't glued to your phone that weekend, here’s the gist: the OpenAI board fired Sam Altman out of nowhere.
It was chaos.
They claimed he wasn't "consistently candid," but nobody really knew what that meant. Then, in a move that felt like a movie script, nearly the entire company threatened to quit unless he was brought back.
The public gratitude tour
When Sam finally returned to his seat as CEO, he didn't come back swinging. Instead, he went on a massive gratitude tour. In a memo that was eventually shared with the public, he used the words "thank you" more times than a bridesmaid at a wedding.
He specifically thanked:
- Mira Murati: For stepping up as interim CEO and keeping the ship steady.
- Greg Brockman: His co-founder who quit in solidarity the second Sam was ousted.
- Satya Nadella: The Microsoft CEO who basically offered to hire the entire company overnight.
- The Employees: Who signed a letter saying "OpenAI is nothing without its people."
It was a masterclass in corporate diplomacy. Even towards Ilya Sutskever—the man who was widely seen as the architect of the coup—Sam was incredibly graceful. He said he loved and respected Ilya and harbored "zero ill will."
📖 Related: Why the Heart Eyes Emoji iPhone Look Keeps Changing
Does Politeness Actually Make ChatGPT Smarter?
There’s this persistent rumor in the AI community that being nice to the model actually gets you better results. You've probably seen the "tips" on Reddit or Discord: "Tell the AI it's doing a great job" or "Offer it a $20 tip for a perfect answer."
Technically, the model doesn't "want" a tip. It doesn't "care" if you're nice. However, there is a nuance here. LLMs are trained on human data. In the training data, polite requests are often followed by high-quality, helpful responses. By using a polite tone, you might be nudging the model into a specific "persona" or region of its latent space that is associated with being thorough and professional.
So, while saying Sam Altman thank you might cost OpenAI millions in server fees, it might actually be helping you get a slightly better email draft. Sorta.
The Ethics of Wasted Energy
We can't talk about this without mentioning the environmental side of things. In 2026, we’re more aware than ever of how much energy these AI models gulp down. If we know that "please" and "thank you" are burning through millions of dollars in electricity, should we stop?
Some critics argue that in a world facing a climate crisis, wasting energy on "digital manners" is a luxury we can't afford. Others think it’s a drop in the bucket compared to the massive training runs required for models like GPT-5 or Sora.
👉 See also: The Pirate Bay Torrent Survival Story: Why It Just Won't Die
The Trade-off
- The Pro-Politeness Side: Maintaining our own humanity and habits is important. If we start being jerks to AI, that behavior might bleed into how we treat real people.
- The Efficiency Side: We should treat AI like a tool. You don't say "thank you" to your hammer or your toaster. Efficiency should be the priority.
Actionable Takeaways for AI Users
If you want to be a more "efficient" AI user without feeling like a total jerk, here are a few ways to handle your prompts moving forward:
- Focus on the Goal: Instead of "Please could you kindly write me a report," just start with "Write a report on..." The AI won't be offended.
- Batch Your Praise: If you really want to express gratitude, do it at the end of a long thread rather than after every single message.
- Use System Prompts: If you’re using the API or Custom GPTs, you can set a "polite" tone in the instructions so you don't have to type it every time.
- Mind the Tokens: If you’re working on a complex task where every token counts (like coding), skip the pleasantries to save context window space.
At the end of the day, Sam Altman is clearly okay with the bill. Whether it's the millions spent on polite ChatGPT users or the public "thank you" notes that saved his company, gratitude seems to be a core part of the OpenAI brand. Just maybe don't feel too guilty the next time you forget to say thanks to the bot—it's probably doing its wallet a favor.
Next Steps for You:
Check your own ChatGPT history. Look at how many tokens you might be "wasting" on pleasantries. If you're using AI for business, try stripping your prompts down to the bare essentials for one week and see if it changes the quality of the output. You might find that the AI is just as helpful without the "please," saving both time and a little bit of the planet's energy.