Europe is investing tens of billions of euros in AI Factories, AI Gigafactories, and Neoclouds. This is the computing infrastructure that is expected to power the next generation of artificial intelligence and turn Europe into an “AI Continent”. Much of the discussion has focused so far on GPUs, high-performance networking, and the advanced storage these facilities will need. These technologies are essential, of course, but they are only part of the equation. The software that orchestrates these new AI infrastructures is becoming just as strategic as the hardware itself.

This is no longer only a future policy direction. The recently published EuroHPC call for AI Gigafactories requires proposals to include credible plans for the progressive integration of EU-based software solutions and calls for a “European-based and controlled software stack (cloud and AI)” to become operational, where possible from Phase 1, and in any case within four years of operational readiness. The evaluation criteria explicitly favor proposals that reduce dependence on non-European technology providers.

Software Sovereignty Is Becoming a Procurement Requirement

The European Commission’s proposal for the Cloud and AI Development Act (CADA), adopted on 3 June 2026, proposes a single EU-wide framework for assessing cloud and AI sovereignty. The Commission’s existing Cloud Sovereignty Framework offers another important signal from public procurement: sovereignty is being assessed across strategic, legal, data, operational, supply-chain, technology, security, and sustainability dimensions.

Europe’s concept of sovereign cloud goes well beyond the physical location of data or formal GDPR compliance. Taken together, these initiatives consider legal jurisdiction, ownership, and control, operational oversight, privileged access, the residency of operational data and telemetry, exposure to extraterritorial legislation, and technological dependence on non-European providers. Software sovereignty is therefore one component of a broader operational and technological sovereignty model.

This represents a fundamental shift in a technology market currently dominated by commercial products controlled by non-EU corporations, even if based on upstream open source technologies. In that new scenario, the cloud platform orchestrating an AI Factory is not just another piece of the software stack; it becomes a strategic component of Europe’s digital sovereignty.

Organizations investing today in these CapEx-intensive AI infrastructures should therefore ask themselves an additional question before selecting a cloud platform to run their deployments: Will this platform be able to satisfy the strict software requirements expected for the EU’s most demanding sovereignty levels?

AI Factories Need More Than Accelerators

An AI Factory is much more than a collection of GPU servers. It must orchestrate heterogeneous compute resources, storage systems, networking fabrics, bare-metal servers, virtual machines, Kubernetes clusters, Slurm environments, AI inference services, AI training workloads, and multiple geographically distributed data centers through—ideally—a unified control plane preventing operational silos. It must support multiple hardware vendors, diverse AI frameworks, and hundreds of concurrent tenants, while offering a future-proof architecture that helps to integrate new technological innovations easily.

These capabilities depend entirely on the software platform chosen to run your AI infrastructure. Selecting the right AI management platform is becoming one of the most important architectural decisions for any AI Factory or Neocloud.

Open Source Is Essential—but Not Enough

Open source provides transparency, auditability, interoperability, and freedom from vendor lock-in. However, open source alone does not automatically provide technological sovereignty. Many software components used by European AI Factories and Neoclouds are open source, but in reality their governance, legal control, intellectual property, and commercial ecosystem are based outside the European Union. That’s why many organizations considering adopting an open source solution are now starting to look beyond the obvious:

• Who governs the project and appoints its decision-makers?
• Under which jurisdiction are the governing and supporting entities established?
• Who controls releases, security updates, and the product roadmap?
• Is the software independently operable without a foreign SaaS or licensing service?
• Could extraterritorial laws affect operational data, support access, or continued service?
• Is there a credible European engineering and support capability?
• Are the interfaces, data formats, and deployment architecture sufficiently open to allow substitution?

These factors are becoming increasingly important for publicly funded AI infrastructures intended to support government agencies, critical infrastructure, and strategic sectors such as healthcare, defense, and finance. 

Openness to Global Technology Without Strategic Dependence

A related consideration arises when Neocloud or full-stack AI platform providers license their software to European AI infrastructure projects. Their participation can bring valuable technology, capacity and commercial experience, and the AI Gigafactories call explicitly allows specialised cloud providers to participate. However, consortia must still assess whether dependence on a non-EU-controlled management platform is compatible with their longer-term technological sovereignty roadmap, particularly for the core orchestration and operational layers.

This assessment applies equally to providers established in other trusted European countries outside the European Union, including the United Kingdom. Such providers may be important strategic partners, but they do not automatically qualify as EU-based or EU-controlled technologies for requirements that explicitly use those terms.

OpenNebula: A European Platform for Sovereign AI 

OpenNebula represents a distinctive approach. It is developed and governed by a European company, commercially supported from Europe, and delivered as an open source cloud platform. It provides a unified control plane for bare-metal resources, virtualization, GPU-enabled workloads, networking, storage, elastic Kubernetes environments, Slurm-based AI training, and multi-site cloud operations. It also supports the deployment of LLM inference services, while remaining interoperable with the specialised MLOps, scheduling, observability, storage, networking and AI-service components selected by each operator.

Sovereignty must also be compatible with scale and technological evolution. Next-generation AI infrastructures are expected to reach tens of thousands of accelerators, may span several sites, and may combine different accelerator architectures. OpenNebula supports federation, heterogeneous hardware, phased expansion, and the integration of future accelerator technologies without forcing the operator to replace its control plane.

For organizations targeting the highest levels of software sovereignty, these characteristics are becoming strategic differentiators rather than simply technical features. Organizations making investment decisions today should look beyond GPU performance, networking bandwidth, or cloud features. They should also evaluate who controls the software that will orchestrate their infrastructure over the next decade.

The strategic objective is not technological isolation. It is to ensure that Europe retains effective control over the software layer that governs access to its strategic AI infrastructure. Open, interoperable and European-controlled cloud platforms provide a practical foundation for achieving that objective while preserving access to a global technology ecosystem.

Disclaimer: CADA remains a legislative proposal, and the European Commission’s Cloud Sovereignty Framework was initially developed for specific EU cloud procurement procedures rather than as a universally applicable certification or legal standard. This article discusses the broader strategic direction reflected in these initiatives and in the published EuroHPC AI Gigafactories call. It does not constitute legal advice or a definitive assessment of any platform’s compliance.