Pricing

Pricing is the English word for setting prices. It refers to a company’s decision about what a product should cost. That involves more than a single number. It’s also about exactly what the customer is paying for: per month, per unit, or based on usage. A mobile phone contract illustrates this well: a flat rate for 20 euros a month and a plan with 5 cents per minute can end up costing the same amount, yet they feel completely different. This very choice of billing rule is the core of pricing.

Price determines profit or loss

Price is the only lever a company has that directly brings in money. Everything else costs money first: staff, advertising, servers. That’s why even a small change in price has a strong effect on profit. If a company sells something for 10 euros that costs it 9 euros, 1 euro remains. If it raises the price to 11 euros, the profit per unit doubles.

But prices that are too high come at a cost. Customers switch to the competition or don’t buy at all. So companies look for the point at which volume and margin together yield the most. Nobody knows this point precisely in advance. That’s why pricing is often tested and adjusted later.

With AI services there’s an added twist. Unlike classic software, every single use causes noticeable costs. A text generator has to run expensive graphics chips for every single answer. A flat rate can therefore turn into a loss-making deal if a few users query it extremely heavily. Several providers have already had to revise their pricing models afterward for exactly this reason.

Usage-based, monthly, or outcome-based

In the tech industry, three basic patterns have become established. With a subscription, you pay a fixed amount per month and may use the product within typical limits. With a usage-based model, you only pay for what you actually consume, similar to electricity. With a freemium model, a basic version is free, and the provider only charges money for additional features.

AI providers usually bill usage-based models in tokens. A token is a piece of text, roughly half a word to a whole word. A model then costs, for example, a few euros per million tokens. Anyone building an app can calculate fairly precisely what each user causes in cost. For private customers, though, this is confusing, which is why they usually get a subscription instead.

A common misconception is that price results from costs. In practice it’s often the other way around. Companies first ask what benefit the customer gets, and derive the price from that. This is called value-based pricing. If a program saves a company ten working hours per week, it may cost considerably more than its server bill.

Price tags in the news and in AI products

In business news, pricing almost always comes up in product announcements. When a provider unveils a new language model, the price per million tokens is often the most important figure. Prices for comparable performance have fallen sharply in recent years, in some cases by more than 90 percent. Analysts then speak of a price war, because providers keep undercutting one another.

The topic also comes up constantly in everyday life. Streaming services with ad-supported cheap tiers are practicing pricing. So are airlines, whose ticket prices change hourly because a program monitors demand. This is called dynamic pricing and is controversial among customers. In Germany there are regular debates about fairness and transparency on this topic.

Anyone who tries out an AI application themselves will quickly come across a provider’s pricing page. There are usually several tiers listed side by side, often with a highlighted middle option. This arrangement is no accident but part of the pricing strategy. A very expensive option makes the middle one look like a bargain. Such effects are as much a part of the pricing craft as the plain calculation itself.