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

EuroHPC’s compute infrastructure currently supports mid-sized AI training projects but is not yet capable of handling frontier-class models, according to recent analyses. This limitation is recognized as a key challenge in Europe’s strategic AI development, with the €20 billion InvestAI Facility aimed at scaling capacity through AI Gigafactories.

The EuroHPC Joint Undertaking (JU) has invested €10 billion in supercomputing infrastructure from 2021 to 2027, supporting 19 AI Factories and 13 AI Factory Antennas across Europe. These systems underpin many current European AI projects, including training models like Apertus 70B on Alps. However, the existing compute substrate is operationally sufficient only for mid-sized models and cannot currently support the training of frontier-class models, which require significantly larger and more specialized hardware.

To address this, the EU has launched the InvestAI Facility, with €20 billion allocated to develop up to five AI Gigafactories, each housing over 100,000 advanced AI processors. The goal is to enable the training of trillion-parameter models and support Europe’s ambitions for sovereign AI leadership. The selection process for these Gigafactories is ongoing, with the first projects expected to be operational by late 2026. Meanwhile, flagship systems like JUPITER, LUMI, and Leonardo rank among the world’s top supercomputers, demonstrating Europe’s current technological standing.

Despite these advances, structural issues remain. The current infrastructure exhibits heterogeneity in hardware (CUDA, ROCm, multi-generation GPUs), creating software complexity and optimization overhead for European AI developers. Additionally, flagship systems are geographically concentrated in wealthier member states such as Germany, Italy, Spain, and France, raising concerns about structural inequalities within the EU’s AI ecosystem.

Implications of EuroHPC Infrastructure for Europe’s AI Ambitions

The current EuroHPC compute substrate demonstrates Europe’s capability to support mid-sized AI training but reveals critical limitations for frontier AI models. The €20 billion AI Gigafactory initiative aims to address these gaps, which are essential for Europe’s competitiveness in the global AI race. The infrastructure’s heterogeneity and geographical concentration could influence the equitable development of AI across member states, impacting policy and investment strategies. These developments are crucial as Europe prepares for the August 2026 EU AI Act enforcement window, shaping the future landscape of sovereign AI capabilities.

EuroHPC’s Infrastructure and Strategic Evolution in AI

Since its creation in 2018, EuroHPC JU has coordinated Europe’s supercomputing efforts, with a €10 billion investment in infrastructure as part of the 2021-2027 program. Notable systems include JUPITER (ranked #4 globally), LUMI (#9), and Leonardo (#10). The infrastructure supports numerous AI projects, such as Apertus on Alps and Minerva on Leonardo, which rely on the compute substrate for training models of up to 70 billion parameters. For more on Europe’s AI compute efforts, see Anthropic’s Series H. The recent expansion of the EuroHPC mandate under Council Regulation (EU) 2026/150 explicitly includes AI Gigafactories, signaling a strategic shift toward large-scale AI hardware deployment. Learn more about the future of AI hardware in The Compute Reckoning. However, the current infrastructure’s capacity for frontier AI training remains limited, prompting the EU to prioritize the development of dedicated AI Gigafactories to fill this gap.

“The EuroHPC infrastructure is operationally credible for mid-sized models but structurally insufficient for frontier-class training, which the new AI Gigafactory framework aims to address.”

— Thorsten Meyer

Unresolved Questions About Infrastructure Scalability

It is not yet clear how quickly the AI Gigafactory projects will become operational and whether they will fully overcome the current structural limitations. The procurement process is ongoing, and technological challenges related to heterogeneity and geographical concentration remain unresolved as of mid-2026.

Next Steps for Europe’s Compute Infrastructure Expansion

The ongoing selection and deployment of AI Gigafactories through 2026 will determine Europe’s capacity to support frontier AI training. The June 2026 AI Gigafactory selection timeline and the August 2026 EU AI Act enforcement window will serve as key milestones. Additionally, efforts to address hardware heterogeneity and regional disparities are expected to shape policy adjustments and infrastructure investments in the coming months.

Key Questions

What is the current capacity of EuroHPC systems for AI training?

EuroHPC systems like JUPITER, LUMI, and Leonardo support mid-sized AI models, such as Apertus 70B, but are not yet capable of training frontier-class models at scale.

What are AI Gigafactories, and why are they important?

AI Gigafactories are large-scale facilities designed to house over 100,000 AI processors each, enabling the training of trillion-parameter models and advancing Europe’s sovereign AI capabilities.

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

Heterogeneity in hardware and software, geographical concentration of flagship systems, and capacity limitations for frontier AI training are key structural challenges identified by recent analyses.

When will the new AI Gigafactories be operational?

The selection process is ongoing, with expected operational dates targeted for late 2026, depending on procurement and deployment timelines.

How will the EU address regional disparities in AI infrastructure?

This remains an open question; current efforts focus on expanding the Gigafactory framework and potentially decentralizing infrastructure to mitigate inequalities.

Source: ThorstenMeyerAI.com