EU launches AI Gigafactories call with plan to unlock more than €30 billion for European compute

The planned network could unlock more than €30 billion in combined investment. Up to €10 billion is expected from public sources — split evenly between the EU budget and participating Member States — while the initiative targets at least €20 billion in private co-investment.

The call is a significant step in the EU’s attempt to reduce a structural weakness in the AI market: Europe has strong research, capable companies and demanding industrial use cases, but comparatively limited access to the largest pools of AI compute. The Commission’s announcement frames the Gigafactories as shared infrastructure rather than a single national project, with access intended for businesses, academia and public-sector users across the Union. The Commission’s announcement confirms that applications are now open.

What the EU is actually building

An AI Gigafactory is not simply another data centre. The proposed facilities are designed to support the full lifecycle of advanced AI work: training large models, inference at scale, and fine-tuning systems for specialised European applications.

The programme is structured around two facility categories:

• Up to three larger Category 2 Gigafactories, each eligible for up to €1 billion in EU public funding and planned to house at least 100,000 advanced AI processors.
• Up to four Category 1 facilities, each eligible for up to €500 million in EU public funding and designed around capacity of up to 75,000 advanced AI processors.

For comparison, the real importance of these numbers is not a single chip count but the ability to assemble compute clusters large enough for demanding workloads without requiring a European company or university to build and operate the entire infrastructure itself. The facilities are intended to build on Europe’s existing network of 19 AI Factories, which have provided smaller-scale access to high-performance AI infrastructure.

A public-private model, not a Brussels-only project

The call is aimed at consortia or Special Purpose Vehicles that can combine technical, financial and operational capability. Eligible groups may include technology companies, cloud-service providers, public institutions and private investors. That structure matters: the EU is seeking operators able not only to acquire hardware, but also to run energy-intensive computing infrastructure, provide user access and maintain it over time.

Compute-time procurement will be managed jointly by the European High Performance Computing Joint Undertaking (EuroHPC JU) and participating Member States. Eighteen EU countries have signed the joint procurement agreement that underpins the funding and co-procurement framework.

The Commission has also established letters of intent with hardware manufacturers including Nvidia, AMD and Qualcomm. These agreements do not remove the practical challenge of obtaining advanced processors in a globally competitive market, but they are intended to improve Europe’s route to the hardware needed for the programme.

The investment arithmetic: why €30 billion is a meaningful threshold

The headline figure is large, but the financing structure is equally important. With up to €10 billion in public funding and at least €20 billion from private partners, the target implies a minimum 2:1 private-to-public investment ratio. Put simply, every €1 of planned public funding is expected to be matched by at least €2 from the private sector.

If the programme reaches the maximum of seven facilities, dividing the public envelope evenly would imply an average of roughly €1.43 billion of public funding per site. That is only an illustrative calculation, not the actual allocation model: the official caps are €500 million or €1 billion in EU funding depending on category, while Member State and private contributions will differ by project.

The distinction is useful because it shows the scale required. These are not modest cloud credits or research grants. They are infrastructure projects involving land, power, networking, cooling, hardware procurement and long-term operating capacity.

What this changes for European AI companies

For European start-ups and scale-ups, access to large compute capacity can determine whether they can develop proprietary models, adapt open-weight models, or deploy AI services at meaningful scale. Until now, many ambitious teams have had to depend heavily on non-European cloud infrastructure or overseas model providers.

The Gigafactory initiative does not prohibit the use of foreign AI services. Instead, it creates a potential European alternative for workloads where data location, contractual control, security requirements or procurement rules matter. This is particularly relevant in sectors such as manufacturing, healthcare, public administration, energy, finance and defence-adjacent supply chains.

That is the practical European angle: a company handling sensitive industrial or citizen data may need more than model quality and price. It may also need a clear answer to where data is processed, who operates the infrastructure and how compliance obligations are documented. The Gigafactories are intended to make those choices more realistic for organisations that need high-end AI capacity within Europe.

AI Act implications: capacity alone is not compliance

The timing is notable. The tender opens days before the EU AI Act’s stricter enforcement phase begins in August 2026. For companies deploying interactive AI systems, transparency duties will increasingly require clear user notification when people interact with chatbots and detectable, machine-readable tagging for AI-generated content where applicable.

European compute infrastructure will not automatically make an AI system compliant. A model hosted in the EU can still create legal, security or governance problems if its operator cannot document datasets, manage risks, label generated content or provide appropriate human oversight. But localised capacity can make compliance operations easier where organisations need stronger control over data flows and technical logs.

For teams comparing cloud options, the key question is therefore not only “Where can we get GPUs?” It is also “Can we operate this AI service with the evidence, controls and transparency expected in Europe?”

Timeline: what happens next

The application deadline is 12 November 2026. The Commission expects contract awards in early 2027, with facilities targeted to become operational within 18 months of contract signing.

That means the announcement is not an immediate increase in available compute. European users cannot yet book Gigafactory capacity. However, the procurement call turns the plan into a concrete competition for projects, locations and investment partners — and expands the original ambition from four or five sites to as many as seven after an expression-of-interest phase attracted 76 responses from consortia.

For readers following the wider AI infrastructure race, the development is worth watching alongside the models themselves. Current frontier systems may capture attention, but access to compute influences who can train, adapt and deploy the next generation. Our AI Arena coverage focuses on how model performance translates into real-world use; Europe’s Gigafactory programme addresses the infrastructure layer required to keep that work viable on the continent.