Credit System

A credit system is a kind of internal balance. The customer pays money once and receives in return a certain amount of points, called credits. Every action within the service costs a fixed number of these points. A short text might cost one credit, a generated image twenty, a short video several hundred. Once the balance is empty, you have to buy more or wait until the next month. The principle is familiar from prepaid phone cards or tokens at an amusement park: you exchange real money for a house currency and then spend that house currency.

Why providers don’t simply charge a flat monthly fee

With AI services, the cost per user varies enormously. One occasional user asks ten questions a month. Another has thousands of texts generated automatically. Serving both at the same fixed price would be risky for the provider, because computing power in the data center costs real money. Credits therefore tie the price to actual usage.

For the provider, this has a second advantage. It can change the price of individual features without touching the subscription price. If a new, expensive video model is introduced, it simply gets a high credit price. If computing power becomes cheaper, the credit cost is lowered. This appears less dramatic than a price increase to the subscription itself.

For customers, this is a double-edged deal. Those who use little pay fairly. Those who use a lot easily lose track, because the connection between credits and euros is no longer directly visible. Critics therefore speak of a deliberate obscuring of the real price.

From euro to point and back again

Technically, the provider maintains a counter for each account. Upon purchase or at the start of the month, it is set to a starting value. With every request, the system first checks whether there is enough balance available. The task is then carried out and the appropriate amount is deducted. Some systems only settle the bill after the response, because the costs are not precisely known beforehand.

How many credits a task costs usually depends on the computing time consumed. For text models, this is often counted in tokens, i.e., word fragments that make up the input and the response. A long question with a long answer costs more than a short one. For images, resolution and the number of computation steps count, and for videos, the length of the clip is added as well.

It is important to distinguish this from direct per-token billing, as used by developer interfaces. There, the price is stated in cents and is thus immediately comparable. Credits are an intermediate layer on top of that. Another common misconception is that credits are always valid indefinitely. Many providers let monthly allotted credits expire at the end of the month, while purchased packages remain valid longer.

Credits in subscriptions, no-code platforms, and quarterly figures

Credit systems are most visible in image and video tools. Services like Midjourney, Runway, or Adobe Firefly work with a balance that is included in the subscription and can be topped up. Coding assistants and so-called no-code builders, which let you assemble apps without programming knowledge, also often bill in credits.

In business news, the term surfaces for a different reason. Credits that have been sold but not yet used are, for a company, money in the bank for now, but not yet revenue on the balance sheet. Analysts therefore look closely at how many credits were issued and how many were actually redeemed. A large pile of unused credits can mean that customers are buying more than they need.

In practice, it’s worth doing a quick calculation before signing up. Divide the monthly price by the included credits and estimate how many images or requests that amounts to. Often it then turns out that a cheap plan is exhausted after just a few days of intensive use.