Sovereignty Is a Practice, Not a Podium

The featured image and the post have been generated by Mistral & the podcast has been generated by Google`s NotebookLM

EN

DE

NL

 

One week, two countries, two digital futures

Last week, German Chancellor Friedrich Merz stood at the Digital Summit and called digital sovereignty an “existential question.” Not make-or-buy, he said — who sets the rules. A strong line. A correct line, even. From the hard- to the software, Europe must produce and steer self-determinedly.

One robot handshake and one guided tour later, asked about the German intelligence services planning their new classified cloud with Google as the project partner, the same Chancellor said: Google invests in Germany and Europe too, and “we can enable European sovereignty with an American provider.” And, on the kill-switch concern — the scenario where a US provider can be ordered by its own government to switch your infrastructure off: “I have high trust in our services that the data we have there is secure and protected, even if an American provider builds a large cloud infrastructure, but that one is secure for us.” (Quotes via heise, 03.11.2026; translated from German.)

Read that again. The existential question was answered with a trust declaration. No assessment framework, no BSI criteria catalogue, no exit plan. Trust.

Meanwhile, 200 kilometers west, the same week looked very different.

What sovereignty looks like when someone actually wants it

The Netherlands have been acting for two years on the lesson Germany merely noted: in early 2025, after US sanctions against the International Criminal Court in The Hague, the ICC’s chief prosecutor reportedly lost access to his Microsoft services. Microsoft disputes the account. The Dutch Parliament didn’t care about the dispute — it cared about the dependency. The conclusion: if Washington or Redmond decides otherwise, the lights can go out in The Hague.

Here is what a country does when it draws that conclusion seriously:

  • It blocks the takeover. In May 2026, the Netherlands prohibited the €100m acquisition of Solvinity — the company operating the cloud infrastructure behind DigiD, the national digital identity through which citizens access tax, health and pension records — by US firm Kyndryl. First time ever the Dutch investment screening office stopped a US takeover on public-interest grounds. Reason: the US CLOUD Act would put millions of citizens’ data under US jurisdiction, outside GDPR protection. Kyndryl called it “politicization.” The Dutch called it an objective review that applies to every investor.

  • It builds the alternative. DAWO — Digitaal Autonome Werkomgeving Overheid: a fully government-owned digital workplace, operating system to office suite to cloud to AI, built on NixOS, a Linux that belongs to no corporation (openSUSE and Fedora were rejected because of their corporate ties to US companies). Eight municipalities are piloting; a stable version is due next year.

  • It stops the rollout that fails review. The Dutch tax and customs administration halted its planned Microsoft 365 cloud migration after its ICT review board found deficits in security, future viability and controllability — including a missing exit strategy and insufficient end-to-end encryption in Teams calls. Email and calendar move back on-premises by 2027; European open-source collaboration tools by 2027/28.

  • It buys European. Not even German — actually, precisely German. In April 2026, the Dutch Ministry of Justice and Security officially admitted STACKIT — the sovereign cloud of Germany’s Schwarz Digits — as a cloud alternative for Dutch government agencies, after a full data protection impact assessment.

That last bullet is the one that should sting in Berlin. The German state chooses Google for its intelligence services, while the Dutch government certifies a German sovereign cloud for its own administration. Germany exports digital sovereignty to countries that practice it — and doesn’t import it itself.

The pattern is the scandal, not the decision

To be clear about what this post is not arguing: choosing a US hyperscaler is not inherently incompetent. Under real pressure, pragmatism is legitimate — sometimes you pick a provider and die one death, for now. I have argued before for “EU-based, not EU-only”: ideology is a bad procurement officer.

But pragmatism has a price of admission, and that price is homework:

  1. Name the kill-switch scenario. No US hyperscaler cloud — not even an air-gapped one — runs for long without vendor updates. When the update chain is controlled abroad, “autonomy” applies to your operating state, never to your system’s evolution. The Dutch review board asks these questions before rollout. The German Chancellor answers them with “I have high trust.”

  2. Fund the BATNA. The alternative inventory costs afternoons, not years. My own swarm project found a dozen EU options in a single research session. The German approach is to note the dependency and not even assign someone to look for alternatives. That is not pragmatism. That is convenience wearing pragmatism’s coat.

  3. Make the exit strategy an adoption criterion. The Dutch stopped a rollout because the exit plan was missing. That is the discipline every mid-sized company can copy on Monday morning — and it is exactly what a governance-first AI or cloud strategy means in practice. If you cannot say under what conditions you would leave a provider, you haven’t chosen that provider. You’ve been captured by it.

Germany’s problem is not knowledge. The BSI works on criteria catalogues; the debate is sophisticated; the Chancellor’s speechwriters clearly understand the stakes — the speech proves it. Germany’s problem is that decision precedes review. Trust replaces verification. The podium replaces the practice. And the gap between what the country says and what its institutions do widens, week by week, until the speeches are worth nothing — because everyone, allies and adversaries included, now knows the answer to “who sets the rules?” In Washington, they must be relieved.

What your company should take from this

Sovereignty is not a geography, and it is not a speech. It is a checklist:

  • Do you know which of your critical workflows would stop if a foreign provider pulled the plug tomorrow?

  • Do you have a documented, dated, reviewed exit strategy for your core platforms — before the next migration?

  • Has anyone verified your “we trust our vendor” with an actual assessment, or is it a Merz?

The Netherlands answered these questions with an investment-screening block, an open-source build program, a stopped rollout, and a certified German cloud. Germany answered with a robot handshake and a trust declaration, in the same week, in the same Europe, under the same threat picture.

Two countries. One week. One of them practices sovereignty. The other one podiums it.

Sovereignty is a practice, not a podium. Your move, Berlin.

Trust without scrutiny is always a gamble played with the money of those who weren’t at the table. — Merz in the cloud, SMEs on the stack, the same anatomy.

Sources:

A list of European alternatives

EU Stack Catalogue, Rob 3 November 2026. Dual-use: (a) Procurement matrix for our own organisation, (b) client inventory template / pitch asset for CAIO mandates. Principle: EU-based rather than EU-only — genuine gaps are highlighted, not overwritten. All entries: as at November 2026; verify individually before use (prices, hosting details and ownership structures are subject to change). Category logic: slot first — each slot represents a procurement milestone; the US reference merely indicates what would need to be replaced.


EU-Based Stack — The Catalogue of Alternatives

Method (the same for both purposes): Inventory by slots rather than by brands. For each slot: What is currently in use (US reference)? Is there an EU option? If not — flag this honestly and apply the router logic (a conscious choice with a reason for exclusion rather than a silent default). This is the BATNA search as an afternoon task, carried out systematically once rather than repeatedly during procurement stress.

Models & Inference

Slot US-Referenz EU-Option Herkunft
Frontier-LLM (API) OpenAI, Anthropic Mistral 🇫🇷 Frankreich
Self-hosted LLM OpenAI-Modelle Mistral Open Weights auf EU-Host 🇫🇷 / self-hosted
Reasoning-Frontier o/GPT-Klasse — ehrliche Lücke, Router-Logik —
Übersetzung Google Translate DeepL 🇩🇪 Deutschland
Enterprise-LLM — Aleph Alpha 🇩🇪 Deutschland

Cloud, Hosting & Infrastructure

Slot US-Referenz EU-Option Herkunft
Public Cloud (souverän) AWS, Azure, GCP STACKIT 🇩🇪 Deutschland
Public Cloud (Preis/Leistung) AWS, GCP Hetzner, IONOS, OVHcloud, Scaleway, UpCloud 🇩🇪🇫🇷🇫🇮
Webhosting US-Hoster Alfahosting 🇩🇪 Deutschland
Dateien & Sync Dropbox, Google Drive Filen, pCloud, Tresorit, Koofr, Nextcloud (self-hosted) 🇩🇪🇨🇭🇸🇮

Office, Collaboration & Project

Slot US-Referenz EU-Option Herkunft
Office-Suite Microsoft 365 LibreOffice, Collabora 🇩🇪🇬🇧
Dokumente & Notizen Notion, Google Docs CryptPad, Anytype, Joplin 🇫🇷🇩🇪
Projektmanagement Jira, Asana OpenProject 🇩🇪 Deutschland
Kommunikation Slack, Teams Nextcloud Talk, Jitsi, Element, Wire 🇩🇪🇳🇱🇨🇭
Workflows & Integration Zapier, Make n8n (self-hosted) 🇩🇪 Deutschland
Whiteboarding Miro, FigJam Draw.io, Excalidraw open source

E-Mail, Newsletter & Web-Utilities

Slot US-Referenz EU-Option Herkunft
Newsletter Mailchimp CleverReach 🇩🇪 Deutschland
Formulare Google Forms, Typeform Prifobu Forms, WPForms 🇩🇪 Deutschland
URL-Shortener Bitly urlr.me 🇫🇷 Frankreich
QR-Codes US-Generatoren EuroMetrics 🇵🇹 Portugal
CAPTCHA reCAPTCHA ALTCHA 🇨🇿 Tschechien
Analytics Google Analytics Plausible, Matomo, Umami 🇫🇮🇫🇷

Video & Presence

Slot US-Referenz EU-Option Herkunft
Video-Hosting (Heimat) YouTube, Vimeo Selbst-Hosted auf Alfahosting (DE) 🇩🇪 Deutschland
Avatar-Video HeyGen — ehrliche Lücke, Router-Logik —
Suchmaschinen-Sicht Google Qwant, Ecosia, Startpage 🇫🇷🇩🇪🇳🇱
Deep Research Perplexity, NotebookLM — ehrliche Lücke, Router-Logik —

Security & Identity

Slot US-Referenz EU-Option Herkunft
Passwort-Manager 1Password, Bitwarden KeePassXC, Passbolt 🇩🇪🇫🇷 open source
Verschlüsselung (Files) US-Crypto-SDKs Cryptomator, Tresorit 🇩🇪🇨🇭
Sichere Mail Gmail, Outlook Proton, Tuta 🇨🇭🇩🇪
Security-Keys YubiKey Nitrokey 🇩🇪 Deutschland

Payments

Slot US-Referenz EU-Option Herkunft
Payment Stripe Mollie, Adyen, Klarna 🇳🇱🇸🇪

The honest gaps — and why they make the point

  • Reasoning frontier: no EU model at o/GPT level — router answer: targeted use, documented exit option, multivendor lowers switching costs.
  • Avatar video (HeyGen): no EU equivalent at Avatar-IV level — conscious exception: the transcript (the actual asset) stays vendor-independent; only the rendering is outsourced.
  • Deep research (Perplexity/NotebookLM): no EU counterpart — research hygiene over origin ideology, but with the alternatives inventoried so that switching day is an afternoon, not a project.

Whoever has the inventory negotiates. Whoever doesn’t, asks.