On my journey to FIRE(FInancial Independence, Retire Early) I have been quite mindful of subscriptions. Any recurring payment is designed to make it feel like a small amount in the short term while it effectively is a large amount.

Marketers use the psychological weakness of the average person’s inability to zoom out and see the larger picture to their advantage.

Examples of recurring payments in the descending order of price and necessity are - Rent/Mortgage, Transport/Car payments, insurance, internet and telecom, streaming subscriptions, Fitness/activity club memberships.

The standard subscription playbook is as follows. Start with free -> get people hooked -> Offer a low priced subscription -> offer premium tiers -> Keep increasing the prices.

The last step, i.e. being able to increase prices without losing any subscribers is the holy grail of the subscription pricing model in marketing.

Take a popular subscription service like Netflix for example. It started with just a single price plan. Then they offered premium tiers for 4K, higher number of devices etc.

But, when they tried to increase their subscription prices across tiers by a few dollars, there was a huge uproar. People were livid about Netflix’s greed. Netflix had to buckle under this pressure and came up with various modifications including a lower priced ad supported subscription plan.

A new line item

Now a brand new entrant has come up in the list of subscriptions. AI service subscriptions from companies like OpenAI, Anthropic and Google. They have implemented the subscription playbook in the standard way.

BUT, there is a very important difference. The price difference between the top most tier and the bottom most tier.

For Netflix, this difference is in the range of 3X.

However, for AI services it is in the order of 20X !!!!

Where is the uproar? Why are people not crying out against greed???

Value > Price

The reason no one complains is because the value these AI services provide is greater than the price.

This Value is understood implicitly in the minds of the subscribers as Value = benefit derived minus the amount paid. As long as the perceived Value is positive, subscribers are willing to pay (or at least that is the assumption)

As the models get better the benefit derived increases. This leaves room for subscription prices to increase. This means AI companies can keep increasing the subscription prices almost linearly with the benefits they provide until the Value is positive, even if just by a little bit.

For example, imagine an AI customer service agent which is able to reduce a customer service center’s head-count from 1000 Human Agents to 10 Human Agents.

Let’s assume one human agent CTC (outsourced developing country prices in USD) of 10,000$s annual per agent.

That is about 9.9Million$s in benefits per year (900 Agents at 10,000$s a year). At the limit, companies would be willing to pay an annual subscription of 9.9M$s for that AI service.

Price War?

If the humans fight back through lower pricing in this scenario, then the AI companies can reduce the pricing all the way down to their costs.

It is unclear right now as to what those costs are. AI companies are deep in their CapEx cycle and are burning cash. But these investments are long-lived and will be amortized over time with ever increasing revenues and accrued efficiencies.

This is reflected in the rapid decrease in the $s per token metric and also an almost corresponding increase in the intelligence or value in the tokens generated.

Not to be a Doomer Max but humans have almost no negotiating power, especially in low to medium skilled white collar jobs.

In fact, it can be argued that companies would rather prefer having AI agents than humans and would be willing to pay more than the price of having human employees.

Mainly because having human employees comes with a long tail of labour law litigation risks, training costs, unreliability, disobedience, non-alignment to corporate strategy etc.

A Billion Dollars a month

In 2025, META - the company behind Instagram, Whatsapp, Facebook paid about 120 Billion dollars in total annual salaries and bonuses to its ~80,000 employees.

If Mr Zuckerberg is able to run his company with 1000 top notch employees across domains empowered by AI, then what are the subscription fees he is willing to pay for such a service?

Even a Billion Dollars a month is a steal - 10x lower than human employee cost.

A much more nuanced issue

Take the current top tier subscription prices of frontier models. They are in the ball-park of about 200 USD a month. Even in developed countries, this is not a trivial amount. In the developing world, this is beyond reach.

As the models improve and the pricing increases. The most useful models will also be the least affordable to vast numbers of humans.

In a way, it is only the already wealthy who will be able to access the benefits of AI to become more wealthy. It strangely resembles landownership during our feudal past.

I am curious to see how high could AI subscription prices actually go? At what point would people say the price is too high? Is there a good enough model for the average person?

Interesting times