Vendor Lock-in

Vendor lock-in describes a dependency on a single provider. A company purchases a technology and builds its processes around it. Later, it realizes: switching to a competitor would be possible, but very expensive and slow. So it stays, even if prices rise or the service gets worse. The English term literally means “being locked in by the supplier.” You are not legally trapped, but practically: the door is open, but the way out costs too much.

Why dependency in the AI business becomes so expensive

Whoever has no alternative negotiates poorly. A provider that keeps its customers tied down can raise prices without losing them. This is exactly why investors pay attention to how strongly customers are bound to technology companies. High retention is considered a quality feature of the business model on the stock market. For customers, it is a risk on the balance sheet.

With AI, the effect is particularly pronounced because the sums involved are large. A single AI data center costs billions. Anyone who installs hardware from one manufacturer there cannot replace it after two years. Then there’s the software: programs written for a particular chip family often only run on other chips after months of rework.

However, lock-in is not fraud and not automatically bad. Close attachment to one system also brings advantages: everything fits together, there is a single point of contact, fewer sources of error. It only becomes problematic when the customer failed to factor in the exit costs at the time of purchase.

What exactly the dependency hinges on

The most important glue is data formats. If a provider stores data in a proprietary structure that no other program can read, that data is worthless without elaborate conversion. Interfaces work similarly—the defined ways in which two programs communicate with each other. Every provider defines them slightly differently. Switching means developers have to reprogram every one of these connection points.

A second factor is acquired knowledge. Employees who have worked with a system for years are quick and confident in it. After a switch, they are beginners again, and the company loses months of productivity. These costs appear in no quote, but are often higher than the purchase price of the new software.

Then there are contractual elements: minimum terms, discounts for long-term commitment, fees for exporting one’s own data from a data center. With cloud providers—that is, companies that rent out computing power over the internet—this exit fee in particular is a well-known point of contention. Uploading is cheap, retrieving costs money.

The term in quarterly figures and antitrust proceedings

In business news, lock-in almost always comes up in connection with three topics. First, AI chips: Nvidia supplies not only hardware but also the programming environment CUDA, with which the majority of AI software is written. This software in particular is considered the real reason for its strong market position. Competitors must not only build better chips but also catch up on years of software development.

Second, cloud services. Competition authorities in the EU and the UK have investigated whether exit fees and technical hurdles distort the market. Third, language models: whoever builds their products on one provider’s interface is tied to that provider’s price list. That is why some companies deliberately rely on open models that they can operate themselves.

For you as a user, the same thing happens on a smaller scale. Photos, notes, and messages in a closed ecosystem can only be taken along with great effort. Anyone who has ever switched from one phone operating system to another knows this feeling. The countermeasure is always the same: choose open formats and, right at the start, ask what the exit would look like.