AI, work and moneyMiddleLesson 1845 min read

What AI really costs

Many AI apps feel free. Running them uses expensive chips, electricity and people. Someone pays, and it helps to know who.

✦ Ask the tutor🧪 Lab · 20 min📝 Lesson test
Lesson in motion

In 60 seconds

What AI really costs

Many AI apps feel free. Running them uses expensive chips, electricity and people. Someone pays, and it helps to know who.

1/5
In simple words
A free sample at a sweet shop still cost the shop money. AI apps are similar. Every answer uses powerful computers and electricity, so the company must earn money somehow.
When you ask an AI chatbot a question, your words travel to large buildings full of computers called data centres. Special chips do a lot of maths to produce each word of the reply. That work costs money every single time.

Where the costs come from

CostSimple meaning
TrainingTeaching the model at the start. It needs huge amounts of data, many specialised chips and a long time.
Running (inference)Producing answers for users. Smaller per question, but it happens millions of times.
Chips and hardwareSpecialised processors are expensive and in high demand.
Electricity and coolingComputers use power and get hot, so data centres must be cooled.
PeopleResearchers, engineers, safety teams, support staff and content reviewers.

Tokens: how AI use is often measured

Models read and write text in small pieces called tokens. A token is often part of a word. Many AI companies charge businesses by the number of tokens going in and coming out. Longer questions, longer documents and longer answers cost more.
  1. 1

    You send a message

    Your words, plus any attached document and earlier chat, are counted as input tokens.
  2. 2

    The model works

    Chips calculate the reply, one token at a time.
  3. 3

    The reply comes back

    The reply is counted as output tokens.
  4. 4

    Someone pays

    A business pays per token, or you pay through a subscription, adverts or your data.

Why “free” apps need a business model

  • Freemium: basic use is free, and people pay for more features or more use.
  • Subscriptions: a monthly fee.
  • Advertising: the app shows adverts, sometimes based on what it knows about you.
  • Data: some services may use what you type to improve their products, depending on their terms and settings.
  • Bundling: AI is included to make you buy or keep another product.
Note
None of these models is automatically bad. The key question is: do you understand the deal? Read the privacy terms to see how your conversations may be used and whether you can switch that off.
Careless use
  • Pasting a 100-page document to ask one simple question
  • Assuming a free app has no privacy trade-off
  • Paying for three AI subscriptions you rarely use
Thoughtful use
  • Sharing only the part of the document you need
  • Checking the app's settings and privacy terms
  • Paying only for the tool that clearly saves you time
Real example
Everyday example. A college in Jaipur wanted an AI helper for student questions. The first plan sent every full rule book with every question, and the monthly bill was far higher than expected. By sending only the relevant page each time, they cut the cost a lot. Knowing how tokens work saved real money.
Try this. Open an AI app you use and find its pricing page and privacy settings. Write down: Is there a free tier? What are the limits? Can you stop your chats being used for training? How does the company earn money?

Watch and read more

Lab

A comparison table of how three AI apps make money and treat your data.

~20 min

The problem

Choose three AI apps or services you have heard of.

  1. For each, find the official pricing page and the privacy policy.
  2. Fill a table: free tier (yes/no, limits), paid price, how it earns money, whether your chats may be used for training, whether you can turn that off.
  3. Write down anything you could not find clearly.
  4. Write three sentences: which one gives the clearest deal, and why.

Use only information from the companies' own pages. Note the date you checked, because prices change.

You are done when

Hard questions

Try to answer before you reveal. If you can answer these, you understood the lesson.

Q1Why might an AI company choose to lose money on each free user?Reveal

Because users today can become paying customers later, and a large user base can attract investment, partners and feedback that improves the product. It is a bet on the future. The risk is that if the company cannot find a way to earn enough, it may raise prices, add adverts, change its data terms or close.

Q2Two answers are equally correct, but one is three times longer. Why can the shorter one be better for a business?Reveal

It costs less to produce, is faster to read and is often clearer. Across millions of answers, the saving in cost and time is large. Good design asks for the length the reader needs, not the most words possible.

Please sign in to continue.

Questions people ask

If AI is so expensive, why do companies give it away?

To attract users, learn what people want, build habits, and later sell paid plans or other services. Some companies also accept losses for a while, hoping to lead the market.

Does a longer answer really cost more?

For services that charge by tokens, yes. More words out means more calculation. That is why businesses often ask for shorter answers when detail is not needed.

Will AI get cheaper?

The cost of a given level of AI ability has often fallen as chips and methods improve. At the same time, newer and larger models can be expensive. It is hard to predict exactly how prices will move.

Does using AI use a lot of energy?

Each answer uses some electricity, and data centres use a lot in total. The amount per question depends on the model and the task. The lesson “AI and the environment” in the AI rules and society track looks at this more closely.

Lesson test

5 questions. Get 3 right (60%) to pass and complete this lesson.

Sign in to take the test and save your progress