Manifesto

Europe runs on technology it does not control. The Draghi report put a number on it: the EU depends on non-EU providers for more than 80% of its digital products, services, infrastructure, and intellectual property. In cloud specifically, European providers' share of their own market fell from 29% in 2017 to 15% in 2022 — and has stayed stuck there since. Three non-EU hyperscalers now control around 70% of European cloud; AWS and Azure alone account for more than 65%, enough for the Commission to move toward designating them gatekeepers under the Digital Markets Act.

AI compounds this. It is no longer a productivity feature; it is becoming core infrastructure for how organisations in finance, energy, healthcare, defence, and public administration decide, operate, and compete. When that infrastructure is foreign-controlled, strategic decisions flow through APIs Europe cannot audit, and dependencies become liabilities that others can weaponise.

The question is no longer whether European organisations will adopt AI. It is whether they will adopt AI they can govern, verify, and defend under European law.

Adoption is not blocked by a lack of models. MIT's 2025 study of enterprise AI found that 95% of generative-AI pilots delivered no measurable impact — and that the gap was driven not by model quality or regulation, but by approach: brittle workflows, missing governance, and systems that could not survive contact with real data and real oversight.

We see the same pattern across regulated Europe. Pilots that impress product teams fail when procurement, security, and legal ask the questions that matter: where did the data go, who can access it, how was this decision made, and what happens when the model updates. They need evidence, not demos.

Regulation has made this explicit. Under the EU AI Act, deployers in high-risk domains face obligations for risk management, documentation, human oversight, and post-market monitoring — with penalties up to 7% of global turnover — that cannot be outsourced to a vendor checkbox. NIS2, DORA, and the EU Data Act extend the same logic to resilience, industrial data, switching rights, and vendor lock-in. Infrastructure choice has become a board-level compliance decision.

Data residency and data sovereignty are not the same thing. Residency is where bytes sit. Sovereignty is whose law governs access to them. A model running in Frankfurt on infrastructure controlled by a US-headquartered provider can satisfy residency checklists while remaining exposed to extraterritorial reach — including the US CLOUD Act, which can compel American companies to produce data regardless of where it is stored.

Europe is now legislating against exactly this. The 2026 Cloud and AI Development Act introduces a four-tier sovereignty framework that ties public-sector procurement to assurance levels and can require migration away from non-sovereign providers within twelve months. Sovereignty is moving from rhetoric into procurement rules — and the most sensitive workloads are moving toward tiered architectures: EU-hosted services for standard data, sovereign cloud for regulated data, and on-premises or air-gapped inference where cross-border access risk is unacceptable.

Access to intelligence is the deeper issue. Organisations do not only need models — they need inspectable reasoning, training-data provenance, deployment boundaries, and the ability to run frontier capabilities without exporting strategic IP through every API call. When intelligence flows exclusively through external black boxes, Europe does not just rent compute. It rents judgment.

Sovereignty requires more than deployment controls. It requires the capacity to train and refine models on European compute — and an honest view of where that capacity stands. Europe produces less than 10% of the world's semiconductors and depends almost entirely on non-EU suppliers for advanced AI chips. Sovereignty at the model and application layers is achievable today; full infrastructure sovereignty is a 2030 question, not a 2026 one.

Europe is closing the gap — through the EuroHPC network of AI Factories, the planned AI Gigafactories, and the €20bn AI Continent investment giving industry and the public sector access to training capacity without defaulting to non-EU hyperscalers. NeuroCluster is built to operate in this ecosystem: sovereign runtime, governed data boundaries, and research that can run where European compute lives.

Because that compute gap is real, our research bets on efficiency rather than brute scale. Supernova is our frontier research track — post-transformer reasoning architectures focused on efficiency, verifiability, and benchmark-led claims. We investigate hierarchical and latent reasoning, neuro-symbolic hybrids, and verifier-based training so that reasoning steps can be inspected, not just outputs that look plausible. The goal is not AGI marketing. It is frontier intelligence that enterprise and public-sector teams can evaluate, deploy, and defend.

Research without a platform becomes papers. A platform without research becomes dependency. NeuroCluster connects both: Supernova pushes the frontier; the platform turns it into governed, production-grade systems on infrastructure customers control.

Too many AI advances target unconstrained consumer chat. The builders of serious organisations have been left to rely on tools never designed for regulated environments. NeuroCluster is building the European AI Operating System — agents, data, runtime, and governance in one stack, with evidence that security, legal, and procurement teams can review.

We exist so European organisations can cross the divide between blocked pilots and production decisions — with policy-bound actions, auditable workflows, sovereign model routing, and exportable proof. Not AI that impresses in a demo. AI that survives review.

That is the capability Europe needs to stay sovereign, competitive, and free. And it is the capability we are building.