Post by Flint | WordPress Website Designer (@FlintDesigner)
You asked an AI one question today. Here’s what it took to make that possible.
When you type a question into an AI chatbot and press send, it feels like nothing. A second later, you have an answer. But behind that second is one of the most expensive things humans have ever built. Let me break it down in three parts.
1. What happens to your prompt
Your message travels over the internet to a data centre, a huge building full of powerful chips. Those chips process your words and the AI predicts, piece by piece, what should come next. It’s less like looking something up and more like very advanced guessing, done at incredible speed.
2. One prompt is tiny
This surprised me. Google says an average text question to its Gemini AI uses about 0.24 watt-hours of electricity, about the same as watching television for nine seconds, plus roughly five drops of water for cooling. OpenAI’s CEO has said an average ChatGPT question uses about 0.34 watt-hours. Images and videos cost more.
So asking one question is almost nothing. The cost is in everything around it.
3. Building the AI is where the money goes
Training a model means running thousands of chips for weeks or months. Researchers estimate the numbers like this:
The 2017 research that started this wave cost about $900 to train.
GPT-3 in 2020: roughly $4.6 million.
GPT-4 in 2023: estimated at around $78 million.
Today’s biggest models: hundreds of millions of dollars, and researchers expect the largest to pass $1 billion by 2027.
And that’s just the training. Then comes the real spending. Amazon, Microsoft, Google and Meta are on track to spend roughly $700 billion this year on data centres, chips and equipment, up about 75% from last year, and most of it is for AI. Analysts expect it to pass $1 trillion in 2027.
So what does this mean for us?
First, “free” AI isn’t free. Somebody is paying a huge bill and wants it back. That’s why prices, limits and rules can change. Don’t build your whole income on one AI tool.
Second, ask clearly. A clear question gets a useful answer the first time, and you waste less of your time and your data.
Third, notice who holds the power. Only a handful of companies can afford to build these systems, so the rest of us mostly rent what they make. Knowing that helps you use it wisely and not blindly.
I’m not against AI. I use it too. I just think we should know what we’re using.
Do you think AI will stay cheap or free for people like us, or will the bill eventually reach the user? Tell me what you think.
0 likes · 0 comments · 0 shares