ATAILA Newsroom · Budapest · 2026-08-22

Europe is not missing out on AI. It is missing out on the infrastructure.

When a central-bank president talks about where the gains of a technology wave end up, it is no longer a technology question — it is a question of investment direction. And this is exactly the point where a 20–200-person European company has an actual decision to make: not a lobbying matter, not an EU package, but a contract it either signs or does not.

What portfolio.hu reports

The Hungarian business news site portfolio.hu reports on a speech Christine Lagarde, President of the European Central Bank, gave at the World Economic Forum’s Geneva headquarters. Lagarde’s warning comes in two parts, and the second is the more interesting one.

The first is the historical part: the business gains from the spread of information and communication technologies — in the article’s wording — were captured back then by other regions, not Europe. As the article quotes the ECB president: “We cannot afford to repeat this experience with artificial intelligence.”

The second is the encouraging part: citing survey data, the article reports that euro-area companies are already putting substantial money into the technology.

The article is a short, Bloomberg-based news item, and it stops there. We will not, because between those two sentences sits an unspoken assumption worth saying out loud.

We agree with the diagnosis. The next sentence is missing

The 9% is not a result in itself. It is a flow of money, and a flow has a direction.

If a European company puts nearly a tenth of its investment budget into artificial intelligence, then this year it decides where that money lands. The current default: an American hyperscaler subscription — where not only the bill leaves Europe but also the data that makes the work valuable. The investment genuinely happens, the statistic genuinely improves, and yet exactly what Lagarde fears comes true: the spending is European, the accumulated value is not.

This difference is not ideology. It is entirely practical:

• If the model runs elsewhere, the prompt itself is data processing — and the data has left your control.
• If the platform is elsewhere, the switching cost does not work for you — it works against you.
• If the system stands elsewhere, the operational knowledge builds up with your provider, not with you.

What ATAILA does with this

Where does the compute run, and who holds the data?

ATAILA Cloud is our own EU data-centre capacity — not a hyperscaler resale. The same sovereign environment our own systems run in, under daily production load. Your documents, your embeddings, your logs and the application itself live on infrastructure you control. Data, model and audit log stay in the same hands: yours.

The sovereign EU cloud →

How much of the 9% goes to AI at all?

In our experience: not much. Most of the money goes not to the model but to everything around it — environments, access control, CI/CD, monitoring, backups, security review — the layer a 20–200-person company cannot afford a dedicated platform team for. ATAILA Factory makes that layer repeatable: you describe the application once, and the engine builds, releases and audits it across the environments. One form → a full stack, in production. So the investment goes into what you actually wanted to pay for.

The provisioning engine and the Release Manager →

And what does the team use in the meantime?

Your team already uses AI — the only question is whether it is yours. ATAILA Studio is a managed developer workstation with private AI: code, internal documents and customer data do not end up at a public provider because someone needed a quick answer. Plug in and work; maintaining the machine is our job.

Studio →

Does it work, or does it just sound good?

What we ship, we also operate — not a project handover but a running platform. Our own open*.hu reference products run in production on this platform, and the newest of them reached production in hours through reuse.

Live references →

Meanwhile the market is splitting into columns: some rent out GPUs, some develop, some operate. We carry the whole row — and if you want the value to stay in Europe, the whole row has to stay here, not one of its columns.

Three things we do not claim

We do not claim sovereignty is a return on investment by itself. A bad workflow stays bad on private AI — just more expensive. If the data is not there, or the process is not worth automating, we will say so before you spend anything.

Nor do we claim that every AI workload belongs on a private platform. If your inference needs outgrow this model by orders of magnitude, or you want to train a foundation model, that is not our business: we are not a model provider, and we do not replace the hyperscalers for everyone. But the handful of workloads you cannot put into the public cloud — that is exactly our terrain.

And we do not claim Lagarde’s warning was about our product. A central-bank speech is not a customer endorsement. All we claim is that the logic inside it — the gains end up where the infrastructure is — holds at the level of a 20–200-person company just as it does at continental scale.

Let’s talk about your 9%

If you have that one confidential workflow that fits into this year’s 9% — and that you would not let outside the company’s walls — tell us what you are working on. This is not a sales call: we will tell you honestly whether it can reach production, and what it costs. The monthly fee is fixed; no usage-based bill arrives at the end of the month.

Get started →