📊 Full opportunity report: EuroHPC. The compute substrate. on ThorstenMeyerAI.com — validation score, market gap, and execution plan.

EuroHPC’s compute substrate is operationally supporting Europe’s AI projects, including mid-sized models like Apertus 70B, but faces structural limitations for frontier-scale model training, which the AI Gigafactory initiative aims to address.

The European High Performance Computing Joint Undertaking (EuroHPC JU) has established a significant compute infrastructure through 19 AI Factories across 21 countries, supporting various AI projects and models. Notably, systems like JUPITER, LUMI, and Leonardo rank among the top supercomputers globally, underpinning Europe’s AI research efforts. However, these facilities primarily support mid-sized models, with the largest operational example being Apertus 70B on Alps, demonstrating the infrastructure’s capability at this scale. The ongoing €20 billion InvestAI Facility aims to establish up to five AI Gigafactories capable of training trillion-parameter models, addressing the current capability gap for frontier AI training. The selection process for these Gigafactories is underway, with decisions expected through summer 2026, aligning with the EU AI Act enforcement timelines. The compute substrate faces three key structural challenges: first, the bifurcation between AI Factories and AI Gigafactories highlights that current infrastructure suffices for mid-sized models but not for frontier training; second, hardware heterogeneity and fragmentation—CUDA, ROCm, multi-generation systems—impose significant software complexity and optimization overhead on European developers; third, geographical concentration of flagship systems in wealthier member states creates structural inequalities, risking uneven AI development across Europe. These issues surface as critical limitations that the current infrastructure framework must overcome to realize Europe’s AI ambitions.

Implications of EuroHPC Infrastructure for European AI Leadership

The EuroHPC compute substrate is central to Europe’s strategy for developing sovereign AI capabilities. While it currently supports mid-sized models effectively, its limitations for frontier-scale training pose a risk to Europe’s competitiveness in cutting-edge AI research. The infrastructure’s evolution will influence Europe’s ability to lead in AI innovation, economic growth, and technological sovereignty.

European Supercomputing and AI Policy Milestones

Since its creation in 2018, EuroHPC JU has coordinated Europe’s supercomputing efforts, with a €10 billion investment scope for 2021-2027, including infrastructure and AI Factories. The top supercomputers—JUPITER, LUMI, and Leonardo—rank among the world’s best, supporting various AI projects and models. The current focus is on scaling AI capabilities through the €20 billion InvestAI Facility, which aims to establish up to five AI Gigafactories. These developments are part of a broader European strategy to become a global leader in AI and high-performance computing, with recent policy updates expanding the JU’s mandate to include AI, quantum technologies, and large-scale model training infrastructure.

“The EuroHPC infrastructure framework is the operational backbone for Europe’s sovereign-AI movement, but it faces three structural challenges that could limit its future scalability.”

— Thorsten Meyer

Unresolved Challenges and Future Infrastructure Developments

It remains unclear how quickly the AI Gigafactory procurement process will proceed, whether the planned facilities will fully address the structural limitations, and how hardware heterogeneity will be managed at scale. Additionally, the impact of geographical concentration of flagship systems on broader AI development remains to be fully assessed.

Upcoming Milestones and Policy Decisions in 2026

The AI Gigafactory selection process will continue through summer 2026, with decisions expected by August, aligning with the EU AI Act enforcement timeline. The deployment of these facilities will determine whether Europe’s compute substrate can support frontier AI training at scale. Further, the ongoing integration of heterogeneous hardware systems and efforts to address regional inequalities will shape the continent’s AI landscape in the coming months.

Key Questions

What is the current capacity of EuroHPC for AI training?

EuroHPC supports several high-performance systems like JUPITER, LUMI, and Leonardo, which are capable of mid-sized model training, exemplified by Apertus 70B. However, it is not yet sufficient for training trillion-parameter models at scale.

What are the main challenges facing Europe’s AI compute infrastructure?

The key challenges include hardware heterogeneity and fragmentation, geographical concentration of flagship supercomputers, and the structural gap for frontier AI training capacity.

How will the €20 billion InvestAI Facility change Europe’s AI landscape?

It aims to fund up to five AI Gigafactories capable of training trillion-parameter models, addressing current capacity limitations and reducing dependency on external providers.

When will the new AI Gigafactories be operational?

Procurement decisions are expected by late summer 2026, with operational timelines depending on project execution and deployment schedules.

Will the current infrastructure support Europe’s AI sovereignty?

Partially, for mid-sized models; fully, for frontier AI training, the infrastructure requires expansion through the Gigafactory program and addressing structural challenges.

Source: ThorstenMeyerAI.com