“Can Europe still succeed in the tech sector?” This question is on the trade show’s agenda. In the business world, it takes on a more concrete form: Which tool should we choose to work more effectively, at what cost, and with what guarantees?

In the cloud sector, an imbalance has taken hold. According to Synergy Research Group, Amazon, Microsoft, and Google accounted for 70% of the European cloud infrastructure services market. European providers held about 15%, down from 29% in 2017.

U.S. companies do, in fact, have strong selling points: a wide range of services, significant capacity, and an international presence. For a company eager to launch a product, these advantages carry a lot of weight. Choosing a European provider, therefore, cannot be based solely on its origin. It must also meet the company’s needs, meet deadlines, and remain affordable.

A company may want to reduce its dependence but may not be able to afford to replace its equipment or train its staff.

AI could make us even more isolated

With AI, the same pattern is likely to repeat itself. At first, a team tests an assistant to draft texts or summarize documents. Then it connects it to its files, customer relationship management software, and internal procedures. Gradually, the tool becomes a central part of daily work.

Leaving it behind is no longer just a matter of canceling a subscription. It requires adapting software, verifying that the new tool delivers satisfactory results, and changing employees’ habits. The more AI becomes integrated into the company, the more costly the change can become.

The risk, therefore, goes beyond the issue of data storage. It also concerns work processes and the ability to continue operations if a provider raises its prices, changes its service, or becomes unavailable.

However, this scenario is not inevitable. Comparing several models and avoiding the temptation to build all your processes around a single tool allows you to keep your options open. This precaution should be taken before certain practices become established.

Useful AI must also prove its worth in the financial statements

This potential dependency means we need to look beyond the listed price. The subscription fee is, in fact, only part of the bill. We also need to prepare the data, connect the tools,

Secure access, support teams, and verify responses. Running AI on your own servers adds costs for hardware, energy, and maintenance.

In the 2026 State of FinOps survey, 98% of respondents reported managing AI spending, compared with 31% two years earlier. This figure refers to the technology spending specialists surveyed, not all European companies.

The question, then, is what the company actually gains from this… Because an assistant who produces a response in a few seconds isn’t necessarily cost-effective if an employee then has to spend twenty minutes correcting it.

We must therefore measure concrete results: the cost of a properly processed case, the number of customer requests resolved, or the reduction in errors.

The decision can no longer be left to the IT department

These decisions affect the entire company. Business teams must specify what they expect from the tool. The finance department must calculate its full cost. Security and legal managers must assess the risks and ensure compliance with all applicable requirements. Senior management must decide which dependencies it will accept.

European regulations reinforce this requirement. For the organizations concerned, the NIS2 Directive specifically requires them to take into account cybersecurity risks associated with suppliers. Purchasing a service therefore does not exempt an organization from considering what would happen in the event of an incident at its service provider.

Should we slow down to better protect sensitive data? Pay for a backup solution? Give up a practice whose benefits remain uncertain?

Increasing the number of suppliers isn't an automatic solution either. It can complicate the teams' work and drive up costs. The goal is to reduce the most dangerous dependencies, not to spread out purchases.

Keeping your options open means preparing for your exit

Not all activities require the same safeguards. Drafting public documents and handling confidential client files do not pose the same risks. Efforts must therefore focus first on sensitive data and services whose disruption would paralyze the company.

Before signing a contract or subscribing to a service, asking a few questions can help make data sovereignty a reality. Can you retrieve your data in a reusable format? How long would it take to switch to a different solution? Would your teams know how to do it? Has an alternative been tested?

The Data Act is intended to facilitate these changes. It calls for the elimination of fees for switching data processing service providers as of January 12, 2027. However, technical adjustments and employee training will continue to cost time and money.

European companies therefore still have a choice. However, they may lose that choice—one tool at a time—if they focus solely on immediate gains. Sovereignty is also measured by this ability: the ability to choose a technology because it is useful, and then to replace it when it is no longer useful.