← Blog

How a Dutch Startup Built a Sovereign AI Platform in Under a Year — And What It Signals for EU Enterprises

    A four-person team in the Netherlands built and launched a sovereign AI platform in less than twelve months. The same task, attempted through a hyperscaler, takes EU enterprises an average of 12 to 18 months just to reach production — without the sovereignty guarantees. That gap is not a quirk of bureaucracy. It is a structural feature of how the EU sovereign AI market is forming in 2026, and it has real consequences for every EU enterprise that is still waiting on a vendor contract to start its AI journey.

    Definition

    Sovereign AI platform is an AI deployment where the model, compute, data storage, and operational control all reside within a defined legal jurisdiction — typically deployed on infrastructure owned by an entity with no foreign legal obligations, such as an EU-domiciled company with no US parent. Data processed by the platform is beyond the reach of foreign governments through laws like the US CLOUD Act or FISA, and compliance with GDPR, the EU AI Act, NIS2, and the EU Data Act is built into the infrastructure rather than configured afterward.

    This article maps the actors, the timelines, and the structural forces shaping the EU's sovereign AI platform landscape right now. Whether you are an enterprise architect assessing infrastructure options, an innovation manager trying to get AI approved by your legal team, or a startup founder looking at where the market is heading, the picture is changing faster than most analysts expected.

    Why a four-person Dutch team could solve what €75M EU initiatives struggle to deliver

    GLBNXT was co-founded by Jan Saan, the former CTO of CM.com, one of the Netherlands' most recognized cloud communications scale-ups. The team is small by design. Their sovereign AI platform is ISO 27001 certified and NIS2-compliant, and it targets the specific pain that large EU enterprises experience when they try to deploy AI on hyperscaler infrastructure: the compliance layer takes longer to build than the AI capability itself.

    The reason a small team can move faster than a €75M institutional initiative is not talent or budget — it is constraint. GLBNXT optimizes for exactly one outcome: getting an EU enterprise live on a compliant AI platform as quickly as possible. They are not building a global infrastructure product that must satisfy every regulatory regime simultaneously. They are not managing a consortium of 70+ organizations with competing priorities. They are not waiting for procurement cycles that take nine months on their own.

    This is the same dynamic that has always allowed focused startups to outpace large incumbents in well-defined market segments. What makes the sovereign AI case unusual is that the regulatory complexity — GDPR, NIS2, the EU AI Act, and the EU Data Act operating simultaneously — is precisely what creates the moat. Startups that build compliance in as the default architecture rather than an add-on become structurally faster, not slower, as the regulatory environment tightens.

    62% of European organizations are pursuing sovereign AI due to geopolitical uncertainty, according to an Accenture survey of 1,928 organizations. Only 16% have elevated AI sovereignty to CEO or board-level priority — but that number is changing.

    The demand signal is real. Search interest in "sovereign AI" tripled on Google UK between January 2025 and April 2026. The enterprise conversation has shifted from "should we think about this" to "how do we get this done this year." GLBNXT is positioned at exactly that inflection point.

    The three sovereign AI layers EU enterprises are actually deploying

    Sovereign AI is not a single product decision. It is a continuum across five dimensions: the AI model itself, the hosting infrastructure, training data provenance, operational control, and hardware. In practice, most EU enterprises in 2026 are addressing three of these five layers when they deploy — and the sequence matters.

    Layer 1: Sovereign hosting

    The most immediate and legible layer. Where does the compute live, and who controls it? This is where the US CLOUD Act creates the sharpest legal risk. An EU enterprise that runs AI workloads on AWS, Azure, or Google Cloud is using infrastructure owned by a US-incorporated entity. Those entities are subject to the CLOUD Act regardless of which physical data center stores the data. An AWS Frankfurt region commitment addresses data residency — it does not address data sovereignty. The parent company's US domicile is the legal exposure, not the server rack location.

    Leaf Cloud, a Dutch infrastructure provider, illustrates what genuine sovereign hosting looks like at the infrastructure tier. The company has no parent company, no investors, and no board members outside the EU. That ownership structure — not just a data center location — is what makes it legally beyond the reach of the CLOUD Act and FISA. Leaf Cloud holds ISO 27001 and SOC 2 Type II certification, and is pursuing HAVEN+ certification. It is the kind of foundational infrastructure that sovereign AI platforms like GLBNXT build on.

    Layer 2: Compliant AI application

    The second layer is the AI application stack: models, APIs, data pipelines, and the processing logic that touches customer and employee data. This is where GDPR, the EU AI Act, and NIS2 intersect most directly. Under the EU AI Act, organizations must maintain accountability for AI systems used in their operations — including systems their employees adopt informally. Under GDPR, any processing of personal data by an AI system requires a legal basis, a data processing agreement, and demonstrable controls. Under NIS2, organizations in scope must manage cybersecurity risk across their supply chain, including AI vendors. A sovereign AI platform that addresses all three simultaneously — as GLBNXT claims to do — removes the three-way compliance negotiation that typically adds months to enterprise AI deployments.

    Layer 3: Model and data provenance

    The third layer is harder and less commonly addressed in 2026: what models are you running, where were they trained, and on what data? For most EU enterprises, this layer comes second — after sovereign hosting and compliant application architecture are already in place. But it is the layer that will define the next phase of EU AI Act enforcement, particularly for high-risk AI use cases. The EU's stated intention to develop sovereign and secure AI models specifically within the energy sector signals where institutional pressure is heading. Enterprises in regulated industries should treat model provenance as a near-term compliance requirement, not a future consideration.

    EURO-3C and the EU's institutional sovereign AI push

    Not all sovereign AI infrastructure is being built by startups. EURO-3C is a €75M+ EU-Commission-backed federated cloud and AI initiative, led by Telefónica, with more than 70 participating organizations including telecoms, technology companies, startups, and SMEs. Its purpose is to build a federated European cloud and AI infrastructure that reduces structural dependence on non-EU hyperscalers.

    EURO-3C represents the institutional tier of the EU's sovereign AI push. It is large, well-funded, and designed to create shared infrastructure at a scale that individual commercial providers cannot match. But it is not a product. EU enterprises cannot sign up for EURO-3C the way they can contract with GLBNXT or deploy on Leaf Cloud. The initiative operates on a policy and infrastructure development timeline that is necessarily longer than a commercial startup's deployment cycle.

    The EU is also advancing a 'cloud and AI development act' designed to create a single EU-wide framework for assessing cloud and AI sovereignty, alongside regulatory streamlining for data centre deployment across the EU. A Chips Act 2.0 is planned to build capacity in cutting-edge semiconductor manufacturing — the hardware layer that underpins long-term EU digital independence. These are decade-scale investments that address the structural gap, not the near-term enterprise deployment problem.

    The practical implication for EU enterprises is that EURO-3C and the institutional policy layer create the strategic backdrop, but they do not accelerate your 2026 deployment timeline. Commercial sovereign AI platforms operating on EU-native infrastructure are the only route to production this year.

    The deployment speed gap: 30 days vs 12–18 months

    GLBNXT's claim — that EU enterprises can go live on their sovereign AI platform in 30 days — is worth examining carefully, because the deployment speed gap is one of the most important structural dynamics in the EU AI market right now.

    The 12–18 month figure for hyperscaler AI deployments in EU enterprises is not primarily a technology problem. The AI models exist. The APIs work. The latency is acceptable. What takes 12 to 18 months is the compliance and procurement layer: information security reviews, data processing agreement negotiations, DPIA completion, NIS2 supply chain risk assessments, legal sign-off on cross-border data flows, and the internal change management required to get a new AI vendor past procurement committees. In regulated industries — financial services, healthcare, public sector — each of those workstreams runs on its own timeline, and they rarely run in parallel.

    A sovereign AI platform that is pre-certified for ISO 27001, NIS2-compliant by design, and built on infrastructure that is structurally beyond CLOUD Act reach removes most of those workstreams from the enterprise's to-do list. The legal team does not need to spend three months negotiating GDPR article 28 terms with a US parent company. The CISO does not need to run a separate CLOUD Act exposure assessment. The DPA does not need a special carve-out for AI processing. Those compliance questions are answered by the vendor's architecture before the customer's process starts.

    That is how 30 days becomes plausible. It is not that the technology deploys faster — it is that the compliance surface area is already closed.

    Search interest in "sovereign AI" on Google UK between January 2025 and April 2026, according to Google Trends data. Enterprise demand is no longer latent — it is active and accelerating.

    Shadow AI: the compliance pressure that's driving enterprise demand

    The single most underestimated driver of sovereign AI platform demand in 2026 is not the CLOUD Act, not GDPR, and not NIS2. It is shadow AI: the uncontrolled employee adoption of AI tools that the IT and legal teams did not sanction, did not assess, and cannot audit.

    Shadow AI is the enterprise equivalent of shadow IT, but with a harder compliance edge. When an employee pastes customer call notes into a free-tier ChatGPT session to generate a follow-up email, that is a potential GDPR Article 83 violation. When a sales team member uses a consumer AI meeting tool to transcribe a deal discussion involving a prospect's personal data, that transcript may be sitting on infrastructure subject to the CLOUD Act. When 200 employees across a company have adopted 14 different AI tools — none of them reviewed, none with DPAs in place — the organization's AI Act accountability obligations become impossible to satisfy.

    The EU AI Act does not distinguish between AI your company chose and AI your employees chose. If an AI system is being used in your operations, the accountability obligations attach to you. That is a structural compliance trap, and it is pushing enterprise procurement teams to make a decision they previously deferred: deploy a sanctioned, compliant AI platform before the shadow adoption problem becomes an audit finding or a regulatory incident.

    This is why GLBNXT's positioning — fast deployment of a compliant sovereign AI platform — addresses a real and urgent pain. The enterprise's alternative is not "wait for a better option." The alternative is "watch the shadow AI problem compound while the procurement process runs." For IT leaders who understand that dynamic, 30 days versus 18 months is not a marketing claim. It is the difference between getting ahead of the problem and explaining to the DPA why you did not.

    Read our related guide on what sovereign AI means for EU companies in 2026 for a deeper look at the regulatory framework driving this shift.

    What this means for EU companies evaluating AI meeting and sales tools

    If you are an EU enterprise evaluating AI tools for sales teams, account managers, or customer-facing workflows — meeting intelligence, call analysis, coaching, CRM enrichment — the sovereign AI question is not abstract. It is the first evaluation criterion, not the last.

    Sales calls and customer meetings are among the highest-risk data categories for EU AI Act and GDPR compliance. They contain personal data about prospects and customers. They may contain commercially sensitive information. They are recorded and transcribed, which means they are processed at multiple points in a data pipeline. And they are exactly the use case where shadow AI adoption is most aggressive — because salespeople want AI help, and they will use whatever tool is fastest if the company does not provide a compliant alternative.

    The evaluation framework for EU enterprises considering AI meeting and sales intelligence tools should address five questions in sequence:

    1. Is the infrastructure EU-sovereign? Does the provider run on infrastructure owned by an EU-domiciled entity with no US parent, no US investors on its board, and no structural exposure to CLOUD Act or FISA orders?
    2. Is compliance built in or bolted on? Does the platform hold ISO 27001 and NIS2 compliance as a default architecture, or does it require the enterprise to configure those properties through settings and contract terms?
    3. How long to production? What is the realistic timeline from contract signature to live deployment in your environment, including all security reviews and DPA completion?
    4. Does it address shadow AI? Does the platform give IT and legal visibility and control over AI usage across the sales team, or does it create a new sanctioned tool that coexists with unsanctioned ones?
    5. What is the EU AI Act exposure? For each use case — call recording, transcription, sentiment analysis, coaching recommendations — what is the risk category under the EU AI Act, and does the platform provide the audit trail and human oversight required?

    These are not compliance-team questions. They are strategic questions, because the answers determine whether your AI deployment accelerates your sales motion or creates the regulatory incident that stops it.

    The EU sovereign AI market in 2026 is not waiting for the €75M initiatives to deliver. It is being built, deployment by deployment, by teams like GLBNXT who understood early that compliance speed is a product feature — and that EU enterprises will pay for it. The question for EU enterprises is not whether to pursue sovereign AI. It is whether to be ahead of the wave or behind it.

    See how Numi's sovereign AI meeting intelligence platform is built for exactly this evaluation framework — EU infrastructure, compliant by design, and deployable in 30 days.

    Frequently asked questions

    What is a sovereign AI platform?

    A sovereign AI platform is an AI deployment where the model, compute, data storage, and operational control all reside within a defined legal jurisdiction — typically deployed on infrastructure owned by an entity with no foreign legal obligations, such as an EU-domiciled company with no US parent. This means the data processed by the AI cannot be reached by foreign governments through laws like the US CLOUD Act or FISA, and compliance with GDPR, the EU AI Act, and NIS2 is built into the infrastructure rather than configured afterward.

    How long does it take to deploy a sovereign AI system in the EU?

    EU enterprise AI deployments on hyperscaler infrastructure typically take 12 to 18 months to reach production, primarily due to security reviews, compliance sign-offs, procurement processes, and GDPR data processing agreement negotiations. Dedicated sovereign AI platforms built specifically for EU compliance — such as GLBNXT's offering — claim to get customers live in 30 days by pre-certifying the infrastructure stack and eliminating the compliance integration work that slows hyperscaler deployments.

    What is EURO-3C?

    EURO-3C is a €75M+ EU-Commission-backed federated cloud and AI initiative led by Telefónica, with 70+ participating organizations including telecoms, technology companies, startups, and SMEs. It is designed to build federated European cloud and AI infrastructure that reduces dependence on non-EU hyperscalers. EURO-3C represents the institutional tier of the EU's sovereign AI push — a long-term infrastructure investment, distinct from commercial sovereign AI products available to enterprises today.

    Why are EU startups building sovereign AI faster than hyperscalers?

    EU startups building sovereign AI platforms are faster because they optimize for a single outcome. Hyperscalers build global infrastructure that must satisfy every regulatory regime simultaneously, making EU compliance one workstream among many. EU-native startups like GLBNXT and Leaf Cloud build with EU compliance as the default architecture, not as a later-stage addition. They also move without the internal approval layers that slow hyperscaler product decisions, and without the legal complexity of a US parent company structure.

    How do EU companies avoid the shadow AI compliance trap?

    EU companies avoid the shadow AI compliance trap by deploying a sanctioned, compliant AI platform before employee adoption of unsanctioned tools creates unmanageable liability. The EU AI Act requires organizations to maintain accountability for AI systems used in their operations, including those adopted informally by employees. Deploying a GDPR-compliant, EU-sovereign AI platform with proper data processing agreements, audit logs, and access controls closes the accountability gap before regulators or auditors identify it.

    Numi is the sovereign AI meeting intelligence platform built for EU sales teams. EU infrastructure, 30-day deployment, full compliance.

    Get Early Access