Lock-in (Vendor Lock-in)

Lock-in describes a situation in which a customer is practically chained to a provider. In theory they could switch, but switching would be so expensive or laborious that they refrain from doing so. The English term literally means “to lock someone in”; in German one speaks of Anbieterbindung (provider tie-in). An everyday example: anyone who has spent years collecting photos, messages, and purchases with one phone manufacturer loses quite a bit of that when switching to another system. It is not a contract that holds the customer, but the cost of moving. Experts call these costs switching costs.

Why providers plan for lock-in

For companies, tie-in is valuable. A customer who cannot switch has to stay even when prices rise. This is exactly why many providers sell their product cheaply at the start, or even give it away. They make the money later, once the customer is deeply embedded in the system. That’s why analysts looking at tech stocks pay close attention to how firmly customers are stuck.

For the customer, lock-in is above all a risk. They lose bargaining power because they have no real alternative. If prices rise by thirty percent, gritting one’s teeth is often the only option. Even in cases of quality problems or a data-protection scandal, leaving is not a realistic threat. That’s why large companies check before every contract how difficult a later exit would be.

It is important to draw a distinction: lock-in is not the same as a monopoly. In a monopoly, there is only one provider in the market. With lock-in, there are many providers, but the individual customer cannot get to them. Competition thus only exists for new customers.

What makes exiting so expensive

One reason is data. Anyone with millions of documents stored at a cloud provider — that is, on its servers on the internet — has to retrieve that data when switching. Some providers charge fees for every gigabyte downloaded. The move can take months and cost millions.

A second reason is interfaces. Programs access a service via defined commands, and every provider uses its own commands. If you switch providers, you have to rewrite your own program code. With AI models there is a special case on top of this: instructions given to a model are often painstakingly tuned to that exact model. A different model reacts differently and suddenly delivers worse results.

The third reason is human. Employees have learned a system, workflows are built around it, training has been paid for. This knowledge becomes worthless when switching. That is exactly why providers give away certifications and courses to students. Anyone who learned a tool during their studies will recommend it later on the job.

Vendor lock-in with cloud and AI models

The pattern is most clearly visible with the large cloud providers. Amazon, Microsoft, and Google rent out computing power and storage. The more additional services a company uses there, the more entangled everything becomes. Many corporations therefore deliberately spread their systems across two providers. This strategy is called multi-cloud and costs more, but preserves freedom of action.

The same debate plays out with AI language models. Anyone who builds their products entirely on one provider’s model is dependent on that provider’s prices and rules. If a model is discontinued or changed, the customer has to react. Open models, whose components can be downloaded and run independently, are seen as a counterweight. In addition, there are software middle layers that address multiple models interchangeably.

You often encounter the term in the news in connection with antitrust authorities. The EU regularly examines whether exit fees or closed systems hinder competition. A common misconception, by the way, is that lock-in is always a nasty trap. Tight integration can also make systems faster and more convenient. The real question is not whether tie-in arises, but whether the customer knows the price of it.