Quest 4 of 15
The Economics of AI
Every prompt costs money. Learn training vs inference, token pricing, and how to budget API use—skills that separate casual users from people who can deploy AI responsibly in schools and small businesses. For example, you will calculate how many typical chats fit inside a shared US$1 classroom budget.
Start here
You do not need an economics degree. We break down why AI costs money, what a “token” bill means, and why your school might set usage limits.
Big idea
Building an AI model costs millions upfront (training); every chat message costs a little each time (inference). Providers charge per token, so longer prompts and longer answers cost more.
Learn one idea at a time
Read, explore, then mark each idea when you can explain it.
Idea 1 of 5
Training is the one-time, colossal expense: teaching a model from huge datasets using thousands of specialised chips running for weeks or months. The bill includes the computing time (GPU clusters are expensive to buy and to power), the data work (collecting, cleaning, and filtering billions of documents), and the salaries of scarce specialists. For frontier models, the total runs to hundreds of millions of dollars — spent entirely before a single user opens the app. This is why only a handful of companies train the largest models from scratch, and why most businesses and schools rent access to someone else's model rather than building their own: the economics resemble a power station more than an app.
Live interactive diagrams
Tap nodes, stages, or cards to explore — these diagrams match this module’s ideas.
Training cost vs everyday use
- Huge one-time compute bill
- Pay each time you chat
- Rare for a school to train from scratch
- Classroom budget cares about usage
Tap Left or Right for each example, then "Why?" for the AI explanation.
Choose a deep dive
Open the topics you want to explore. The detail stays folded until you need it.
Deep dive 1Input and output have different costs
The input includes your instructions, pasted notes, and conversation history. The output is the reply the model produces. A long reply often requires more computing than reading a prompt, so a request for a 1,000-word essay can cost more than a request for five revision questions.
Deep dive 2Estimate before you scale
Imagine a study bot serving 200 learners each week, with each learner sending five short questions. A small cost per chat can become a large monthly bill when multiplied by all those messages. Teams set limits, test with a small group, and measure real usage before promising a free service to everyone.
Deep dive 3Worked example: costing a school study bot
Suppose a club builds a revision bot. Each question plus answer averages 800 tokens, and the API charges a small fraction of a cent per thousand tokens. One chat costs almost nothing — but 200 learners × 5 questions × 4 weeks is 4,000 chats a month, and adding pasted revision notes to every prompt could triple the token count. The club's estimate shows the difference between 'basically free for a demo' and 'needs a budget for the whole school'. Their fixes: cap questions per learner, trim the notes to the relevant chapter, and use a smaller, cheaper model for simple definition questions. Costing first turned a vague idea into a plan.
Deep dive 4Why companies give some AI away free
Free tiers exist for business reasons you can reason about: they attract users whose feedback improves the product, they create habits that convert to paid subscriptions, and usage data reveals what to build next. Free tiers also come with limits — message caps, slower models, or terms that allow your inputs to be used for training. Reading what you 'pay' with (attention, data, or eventual subscription) is part of digital literacy. It also explains why a free tool can suddenly change limits or shut down: the money behind it has to come from somewhere.
Deep dive 5Tokens have a price
Providers often charge for input and output tokens. Focused prompts and shorter drafts protect school budgets.
One-minute challenge
Connect this lesson to real life
Name one situation where this idea could help, and one thing a person should still check.
Explore a real-world example
Use the arrows to connect the idea to a visible situation.
Photo example
Example: plan before you spend
Write the task clearly once, then reuse a template, instead of re-pasting long documents every time.
Key terms
Tap a term to flip and read the definition.
Optional further learningFree textbooks and trusted online resources
These sources informed the course structure. Use them to revisit a concept or study it in more depth.
Ready check
Tick each idea only when you could explain it without looking back.
Ready for practice? Try the price-guessing game and budget workshop — estimate costs before you spend real tokens.
Extra context (audience, logistics, curriculum notes)
Built for: Uses real Gemini-style pricing examples; math is simple multiplication—suitable for Forms 3+.
Formats: Worked examples on whiteboard · AI cost estimator chat lab · Budget challenge worksheet
AI Architect Module 4 — training cost, inference, token pricing, budget tracking (AI_Architect_10_Part_Course.md).
Next up
Ready for the next part?
When you've finished the reading, inline exercises, and knowledge check for this part, check the box to continue.