European High-Performance Computing Joint Undertaking
The European High-Performance Computing Joint Undertaking (EuroHPC JU) is a legal entity established in 2018 by the European Union, in partnership with European countries and private stakeholders, to coordinate the development of a federated, secure, and hyper-connected supercomputing infrastructure across Europe, encompassing high-performance computing (HPC), quantum computing, and data ecosystems.[1][2] Headquartered in Luxembourg, it pools investments to procure and operate world-class systems, with a budget of €7 billion for 2021–2027, including contributions from the EU's Digital Europe Programme, Horizon Europe, and matching funds from member states.[1][2] Key achievements include the acquisition and deployment of eight operational supercomputers, such as LUMI in Finland (ranked third globally) and Leonardo in Italy (fourth globally), alongside preparations for the exascale Jupiter system in Germany and the integration of six quantum computers.[2] The initiative supports research innovation, industrial competitiveness, and applications in fields like medicine, climate modeling, and cybersecurity, while recent expansions incorporate AI factories to bolster Europe's technological autonomy amid global competition.[1][2]Establishment and Historical Development
Founding and Initial Mandate (2018)
The European High-Performance Computing Joint Undertaking (EuroHPC JU) was established by Council Regulation (EU) 2018/1488 on 28 September 2018, entering into force 20 days after its publication in the Official Journal of the European Union. This legal framework, based on Articles 187 and 188 of the Treaty on the Functioning of the European Union, created the JU as a body with legal personality, headquartered in Luxembourg, to coordinate efforts among the European Union, its member states, and other participating states in developing supercomputing capabilities. The initiative addressed Europe's lag in high-performance computing (HPC) infrastructure, aiming to reduce reliance on non-European providers and enhance technological sovereignty. The initial mandate focused on developing and maintaining a federated, secure, and hyper-connected world-class supercomputing and quantum computing infrastructure across Europe. Key tasks included acquiring and operating petascale and pre-exascale supercomputers, with targets to deploy pre-exascale systems by 2020 and prepare for exascale capabilities by 2022–2023; supporting research, innovation, and skills development in HPC-related fields such as data analytics, cloud computing, and quantum technologies; and ensuring broad access to these resources for public and private users in research, industry, and the public sector. The JU was tasked with fostering a competitive European HPC ecosystem, including procurement of innovative hardware and software to promote indigenous development and integration of cutting-edge technologies. Financially, the Union committed up to €486 million from the Horizon 2020 and Connecting Europe Facility programs, while participating states pledged at least €476 million in-kind and in-cash contributions, and private members at least €422 million, aiming for a total initial investment exceeding €1.3 billion. The JU's duration was set until 31 December 2026, with administrative costs capped to support efficient operations. Governance was structured around a Governing Board comprising representatives from the Union and participating states, responsible for strategic decisions, procurement, and funding allocation; an Executive Director for day-to-day management; and advisory bodies including the Industrial and Scientific Advisory Board to provide expertise on technological and market developments. This setup enabled coordinated deployment of supercomputers hosted in various member states, emphasizing interoperability and data sovereignty.[1]Expansion Under EuroHPC JU Framework (2019–2023)
In 2019, the EuroHPC JU initiated procurement processes for petascale supercomputers capable of at least 10^{15} floating-point operations per second, alongside preparations for pre-exascale systems, as part of efforts to build distributed high-performance computing capacity across member states.[3] [4] On 5 June 2019, the Governing Board selected hosting entities for precursor-to-exascale supercomputers, approving agreements for sites in Finland, Italy, and Germany.[5] This built on earlier site selections, culminating in eight hosting locations announced that year: Sofia (Bulgaria), Ostrava (Czechia), Kajaani (Finland), Bologna (Italy), Bissen (Luxembourg), Minho (Portugal), Ljubljana (Slovenia), and Barcelona (Spain).[6] Funded by joint EU and national contributions totaling over €960 million for 2019–2020, these procurements emphasized co-design with European industry to enhance sovereignty, though systems primarily integrated components from global suppliers like AMD and Intel.[7] Petascale systems were largely acquired by hosting entities under JU oversight, while pre-exascale procurements were directly managed by the JU. By late 2023, deployments included five petascale machines and initial pre-exascale units, operationalizing capacities from approximately 5 PFlops to over 300 PFlops peak.[8] Key deployed systems during this phase are summarized below:| Supercomputer | Location | Peak Performance (PFlops) | Type |
|---|---|---|---|
| Discoverer | Bulgaria | 5.94 | Petascale |
| Karolina | Czech Republic | 12.91 | Petascale |
| MeluXina | Luxembourg | 18.29 | Petascale |
| Deucalion | Portugal | 9.76 | Petascale |
| Vega | Slovenia | 10.05 | Petascale |
| LUMI | Finland | 539.13 | Pre-exascale |
| Leonardo | Italy | 315.74 | Pre-exascale |
| MareNostrum 5 | Spain | 314 | Pre-exascale |
Recent Expansions and Phase II (2024–Present)
In July 2024, the EuroHPC JU amended its work programme to establish a dedicated AI Factory pillar, integrating high-performance computing with artificial intelligence and quantum capabilities to support Europe's technological sovereignty. This amendment enabled calls starting in September 2024 for deploying AI-optimized supercomputers, with two tracks focusing on new acquisitions or upgrades of existing systems.[10] The first phase of AI Factory deployments began with the selection of seven hosting sites on December 10, 2024, in Finland, Germany, Greece, Italy, Luxembourg, Spain, and Sweden, intended for operational launch in 2025. Five of these sites—Finland, Germany, Italy, Luxembourg, and Sweden—were slated to acquire brand-new AI-optimized supercomputers, while the others would upgrade existing infrastructure to prioritize AI workloads.[11][12] Subsequent expansions accelerated in 2025. On March 11, 2025, six additional AI Factories were selected in Austria, Bulgaria, France, Germany, Poland, and Slovenia, backed by approximately €485 million in combined EU and national funding to provide prioritized access for startups, SMEs, and researchers in developing scalable AI models. By October 10, 2025, six more sites were chosen in Czechia, Lithuania, the Netherlands, Poland, Romania, and Spain under the EUROHPC-2024-CEI-AI-02 call, with deployments planned for 2026 and features like experimental AI model testing platforms in Spain, bringing the total to 19 AI Factories.[13][14] Complementing these core facilities, the EuroHPC JU selected 13 AI Factory Antennas on October 13, 2025, to extend services to national ecosystems, offering training, expertise, and access to AI resources for companies, researchers, and public institutions across participating states, funded by approximately €55 million from the European Commission. Other developments included the United Kingdom's association to the JU in May 2024, enhancing cross-border collaboration, and a September 2025 contract with GÉANT for hyperconnected infrastructure linking supercomputers. Quantum efforts persisted with procurements like the February 2024 tender for a photonic system in France and an October 14, 2025, call for a €4 million Phase 1 Quantum Grand Challenge.[15][16][17][18][19][20]Strategic Objectives and Policy Context
Core Technological Goals
The EuroHPC Joint Undertaking's core technological goals center on establishing a federated, secure, and hyper-connected high-performance computing (HPC) infrastructure capable of exascale performance, defined as at least 10^18 floating-point operations per second, to enable advanced simulations, data analytics, and artificial intelligence applications while prioritizing European-developed technologies for strategic autonomy.[1][2] This includes procuring and deploying supercomputers leveraging indigenous processors such as ARM-based European Processor Initiative (EPI) architectures and accelerator technologies to reduce reliance on non-European hardware, with initial pre-exascale systems like LUMI achieving over 150 petaflops in 2022 and paving the way for full exascale by 2025–2027.[21][22] A parallel objective is the development and integration of quantum computing prototypes to complement classical HPC, targeting hybrid quantum-HPC systems for solving complex optimization and simulation problems intractable for traditional supercomputers, with prototypes expected to demonstrate practical utility in areas like materials science and cryptography by the mid-2020s.[1][22] These efforts emphasize energy-efficient designs, aiming to minimize power consumption per computation—critical given exascale systems' projected multi-megawatt demands—through innovations in cooling, software optimization, and hardware architectures, as outlined in funded projects like SEANERGYS targeting workload-specific energy reductions.[23] Additionally, the goals extend to creating AI-optimized "factories" by integrating GPU-accelerated clusters with HPC resources, supporting large-scale model training and inference for European industries, with recent selections in 2025 adding six such facilities to enhance AI sovereignty and competitiveness against global leaders like the United States and China.[14] This infrastructure is designed for broad accessibility, including small and medium enterprises, to drive applications in precision medicine, climate modeling, and engineering, ensuring scalability and interoperability across a pan-European network.[1]Emphasis on Sovereignty and Competitiveness
The European High-Performance Computing Joint Undertaking (EuroHPC JU) prioritizes technological sovereignty as a foundational objective to mitigate Europe's reliance on non-EU hardware and software providers, particularly amid geopolitical tensions and supply chain disruptions exemplified by U.S. export controls on advanced chips implemented since 2022.[24][25] This drive for autonomy is embedded in the JU's regulatory framework, established under Council Regulation (EU) 2018/1488 and expanded via amendments in 2021, which mandate the development of indigenous computing infrastructures to safeguard critical research and industrial applications in fields like climate modeling and drug discovery.[26] By fostering European-developed processors and architectures, such as those under the European Processor Initiative (EPI) launched in 2018 with €60 million initial funding, the JU seeks to create a resilient ecosystem insulated from foreign vendor lock-in and potential sanctions.[27] Key initiatives underscore this sovereignty focus, including the Digital Autonomy with RISC-V in Europe (DARE) project, initiated on March 6, 2025, with €24 million from the EuroHPC JU to prototype RISC-V-based HPC and AI systems capable of petascale performance.[28][29] RISC-V, an open-standard instruction set architecture, enables customizable, royalty-free designs that bypass proprietary Intel x86 or ARM dependencies, aligning with the JU's Multi-Annual Strategic Programme (2021–2027), which explicitly aims to "contribute to the Union's strategic sovereignty" through technology development.[30] Complementary efforts, such as three research and innovation projects selected in February 2022 totaling €100 million, target independent microprocessors and accelerators to integrate into pre-exascale and future systems, reducing the 80–90% market share held by U.S.-dominated suppliers in Europe's HPC deployments as of 2023.[25] These measures address causal vulnerabilities, including data security risks from foreign-hosted clouds, by prioritizing hyper-connected, EU-controlled federations. In parallel, competitiveness is pursued through sovereignty-enabling investments that position Europe to rival U.S. and Chinese dominance in HPC-driven domains like artificial intelligence and quantum simulation. The JU's strategy integrates AI Factories—19 planned by October 2025, with six additional sites selected that month—to deploy sovereign GPU clusters exceeding 10 exaFLOPS aggregate capacity, supporting the EU's AI Continent Action Plan for global leadership without external dependencies.[14] This builds on pre-exascale procurements, where €2.7 billion in public-private funding since 2019 has accelerated systems like LUMI (Finland, 2022) toward hybrid European-U.S. tech stacks, but with escalating mandates for indigenous components in Phase II (2024 onward) to achieve full exascale autonomy by 2027.[31] Empirical benchmarks, such as Europe's climb to second globally in the TOP500 list by November 2023 with 25% of entries, reflect gains, yet underscore the need for sovereignty to sustain innovation rates against U.S. DOE investments surpassing $1 billion annually.[32] Challenges persist, as evidenced by ongoing hybrid procurements, but the JU's causal emphasis on open standards like RISC-V promises long-term self-reliance, enabling competitive edges in energy-efficient, secure computing for industrial sovereignty.[28]Funding Mechanisms and Budgetary Realities
Sources and Contributions
The EuroHPC Joint Undertaking operates as a public-private partnership, with funding sourced from the European Union, participating states, and private members.[2][1] For the 2021–2027 period, the total budget is approximately €7 billion, enabling the procurement of supercomputing infrastructure, research and innovation activities, and related services.[2][1] The European Union provides €3 billion, drawn from the Multiannual Financial Framework programs: €1.9 billion from the Digital Europe Programme for supercomputer acquisition, deployment, federation, and skills development; €900 million from Horizon Europe for research and innovation; and €200 million from the Connecting Europe Facility for interconnections.[2][1] Participating states, numbering over 30 including all EU Member States and associated countries such as Norway, the United Kingdom, and Türkiye, contribute a matching €3 billion in cash and in-kind resources.[2][1] These contributions primarily support the co-financing of hosted supercomputers, operational costs, and national shares in joint procurements, with host countries typically covering up to 50% of a system's costs alongside EU funds.[2] Private members, including the European Technology Platform for High-Performance Computing (ETP4HPC), the Big Data Value Association (operating as DAIRO), and the European Quantum Industry Consortium (QuIC), are committed to €900 million, predominantly in in-kind forms such as technical expertise, software development, and advisory input for research calls and innovation projects.[2][1] In practice, private sector realizations have lagged projections in initial years; for instance, 2021 contributions from ETP4HPC and DAIRO totaled €2.025 million, reflecting a focus on non-monetary inputs over direct cash funding.[33] A 2024 regulatory amendment expanded the scope to include AI factories, reallocating budgets within existing envelopes to accommodate GPU upgrades and AI model training, without altering core contribution structures.[2] This model emphasizes coordinated public investment to achieve technological sovereignty, though actual private engagement remains contingent on project-specific calls and industry capacity.[2][1]Allocation and Expenditure Patterns
The EuroHPC Joint Undertaking's budget for 2021–2027 totals approximately €7 billion in overall investments, with the European Union's contribution amounting to €1.981 billion from the Digital Europe Programme (primarily for infrastructure procurement and operations), €200 million from the Connecting Europe Facility, and €900 million from Horizon Europe (focused on research and innovation).[2][30] This allocation reflects a pattern prioritizing capital-intensive supercomputer deployments in early years (2021–2024), shifting toward sustained operations, skills development, and specialized initiatives like AI and quantum computing in later phases. Member states and associated private entities cover the remainder through matching contributions, often up to 50% of acquisition costs and varying shares of operations (e.g., up to 50% for quantum systems).[34] Expenditure patterns emphasize procurement and maintenance of high-performance systems, with the Digital Europe Programme funding petascale, pre-exascale, and exascale supercomputers (e.g., two pre-exascale systems like LUMI and LEONARDO reviewed for operational costs in 2023).[35] Horizon Europe allocations support R&I activities, including hardware-software co-design and European Processor Initiative development, representing about 30% of EU funds. Operational expenditures, including energy costs for supercomputers, have prompted targeted grants (e.g., securing commitments in 2023 to cover the Union's share of rising energy expenses). In practice, administrative and staff costs (Titles I and II) show underutilization relative to budgets, while grants (Title IV) and operational procurements dominate implementation.[36]| Category | Indicative EU Allocation (2021–2027) | Primary Focus |
|---|---|---|
| Infrastructure Procurement & Operations | €1.981 billion (Digital Europe) + €200 million (CEF) | Supercomputer acquisition (up to 50% EU share), energy/maintenance; e.g., exascale systems by 2025–2027 |
| Research & Innovation | €900 million (Horizon Europe) | Technology development, software stacks, quantum prototypes |
| AI & Specialized Initiatives | Up to €980 million total EU (incl. €800 million Digital Europe) | AI factories, convergence with HPC; e.g., €70 million for AI Factory Antenna call in 2025 |
| Skills & Access | Integrated in Digital Europe | Competence centres, training; minimum 20% expenditure threshold for certain activities |
Supercomputing and Quantum Infrastructure
Pre-Exascale Deployments
The EuroHPC Joint Undertaking procured and deployed three pre-exascale supercomputers between 2020 and 2023 to bridge the gap to full exascale capabilities, targeting sustained performance above 100 petaflops for advanced simulations in scientific research, while prioritizing European sovereignty in hardware supply chains.[2] These systems, funded through a combination of EU and member state contributions totaling over €500 million for the trio, were selected via competitive tenders emphasizing accelerators like GPUs from European or allied vendors to reduce dependency on non-EU technologies.[8] LUMI, hosted by CSC – IT Center for Science in Kajaani, Finland, employs HPE Cray EX architecture with AMD EPYC CPUs and Instinct MI250X GPUs, delivering a peak performance of 552 petaflops.[41][42] Contracted in 2020 for €200 million, it achieved initial operations in mid-2022 after installation delays related to supply chain logistics, and ranks among the world's top systems for energy efficiency due to its liquid-cooled design powered by renewable energy.[43][44] Leonardo, managed by CINECA in Bologna, Italy, utilizes Atos BullSequana XH2000 with Intel Xeon CPUs and NVIDIA A100 GPUs, providing 250 petaflops peak.[45][46] Awarded in 2020 for approximately €125 million, it was inaugurated in November 2022 and reached full production in early 2023, supporting hybrid CPU-GPU workloads for applications in biomedicine and astrophysics.[46] MareNostrum 5, operated by Barcelona Supercomputing Center in Spain, integrates Bull Sequana XH3000 with Intel Xeon and NVIDIA Grace Hopper superchips, attaining 314 petaflops peak across general-purpose, accelerated, and data analytics partitions.[47][8] Procured in 2022 for €140 million following a retender, it entered the TOP500 at eighth place in November 2023 after deployment in late 2023, featuring direct liquid cooling for enhanced reliability.[48][49]| Supercomputer | Host Country | Peak Performance (PFlops) | Initial Operations |
|---|---|---|---|
| LUMI | Finland | 552 | 2022 |
| Leonardo | Italy | 250 | 2023 |
| MareNostrum 5 | Spain | 314 | 2023 |
Exascale-Class Systems
The EuroHPC Joint Undertaking has prioritized the development and deployment of exascale-class supercomputers, defined as systems capable of exceeding 1 exaFLOPS (10^18 floating-point operations per second) of sustained performance, to address Europe's strategic needs in high-performance computing for scientific simulation, AI, and data analytics.[52] These systems represent a leap beyond pre-exascale platforms, enabling simulations of unprecedented scale in fields such as climate modeling, drug discovery, and materials science, while reducing reliance on non-European computing infrastructure.[53] JUPITER, Europe's inaugural exascale supercomputer, was inaugurated on September 5, 2025, at the Forschungszentrum Jülich in Germany and is operated by the Jülich Supercomputing Centre.[53] Owned by the EuroHPC JU, it utilizes Eviden's BullSequana XH3000 architecture, featuring a modular design with a scalable booster module optimized for compute-intensive workloads and integrated NVIDIA Grace CPU Superchips and HPE Cray EX accelerators for hybrid CPU-GPU processing.[8] [54] Achieving over 1 exaFLOP of performance, JUPITER ranked fourth on the TOP500 list in June 2025, marking a significant advancement in European computing sovereignty.[55] Its deployment supports open-access research, with up to 50% of capacity allocated for peer-reviewed projects across EuroHPC member states.[56] Further enhancements to JUPITER, including a new module installed in November 2024, have bolstered its efficiency and AI capabilities, ensuring sustained leadership in exascale computing amid global competition from systems like Frontier in the United States.[52] While JUPITER stands as the primary operational exascale asset, the EuroHPC JU has outlined procurements for additional systems to reach a total of at least two exascale machines by the mid-2020s, though specific deployments beyond JUPITER remain in planning as of October 2025.[21] These efforts underscore a commitment to indigenous technology stacks, minimizing vendor lock-in through partnerships with European firms like Eviden and Bull.[54]Quantum Computing Prototypes
The European High-Performance Computing Joint Undertaking (EuroHPC JU) has prioritized the procurement and integration of quantum computing prototypes to advance hybrid quantum-classical workflows, aiming to bolster Europe's technological sovereignty in emerging computational paradigms. These prototypes, primarily small-scale systems with tens to hundreds of qubits, are hosted on existing pre-exascale supercomputers across member states, enabling researchers to test quantum algorithms alongside classical high-performance computing resources. Initial selections occurred in October 2022, with procurements focusing on diverse quantum technologies to mitigate risks associated with immature hardware and foster ecosystem development.[57][58] A cornerstone initiative is the High-Performance Computer and Quantum Simulator (HPCQS) project, launched on December 1, 2021, with a €12 million budget co-funded by EuroHPC JU and six member states. Coordinated by Jülich Supercomputing Centre (JSC), HPCQS integrates quantum simulators exceeding 100 qubits each with Tier-0 systems like JUWELS in Germany and Joliot-Curie in France, using tools such as Atos Quantum Learning Machine for hybrid software stacks and cloud access. This federated approach targets practical applications in optimization and simulation, serving as a proof-of-concept for scalable quantum-HPC integration while addressing challenges like qubit coherence and error rates inherent to current prototypes.[59] EuroHPC JU has procured eight such prototypes, emphasizing European suppliers to prioritize domestic innovation over foreign dependencies. These systems employ varied modalities—neutral atoms, trapped ions, photonic, and superconducting—to explore technological viability, with qubits ranging from 12 to 150 and integration on supercomputers boasting petaflop-scale classical capacity. The following table summarizes key prototypes:| Prototype Name | Technology | Qubits | Host Supercomputer (Petaflops) | Location | Supplier |
|---|---|---|---|---|---|
| JADE | Neutral atoms (analog/digital simulator) | 100 data | JURECA DC (23.5) | JSC, Germany | Pasqal (France) |
| Ruby | Neutral atoms (analog/digital simulator) | 100 data | Joliot-Curie (22) | GENCI/CEA, France | Pasqal (France) |
| Lucy | Photonic (digital) | 12 data | Joliot-Curie (22) | CEA, France | Quandela (France) & attocube (Germany) |
| Piast-Q | Trapped ions (digital) | 20 data | ALTAIR (5.9) | PCSS, Poland | Alpine Quantum Technologies (Austria) |
| VLQ | Superconducting (digital) | 24 data | KAROLINA (12.9) | IT4Innovations, Czechia | IQM (Finland) |
| Euro-Q-Exa | Superconducting (digital) | 54 (System 1), 150 (System 2) | SuperMUC-NG (26.9) | LRZ, Germany | IQM (Finland) |
| EuroQCS-Italy | Neutral atoms (analog/digital simulator) | 140 data | Leonardo (315.74) | CINECA, Italy | Pasqal (France) |
| EuroQCS-Spain | Superconducting annealing (analog) | 10/15/25 (systems) | MareNostrum 5 (314) | BSC, Spain | Qilimanjaro Quantum Tech (Spain) |
Specialized Projects and Initiatives
AI Factories and InvestAI Initiative
The AI Factories initiative under the European High-Performance Computing Joint Undertaking (EuroHPC JU) establishes specialized hubs that repurpose supercomputing resources to train and develop trustworthy generative artificial intelligence models, emphasizing data sovereignty and accessibility for European users.[63] These facilities provide free computing power, customized support services, and privileged access primarily to small and medium-sized enterprises (SMEs), startups, and research entities, aiming to foster innovation in sectors such as healthcare, climate modeling, and manufacturing without reliance on non-European cloud providers.[64] The initiative stems from the 2024 amendment to the EuroHPC JU regulation, which expanded its mandate to include AI infrastructure, building on existing pre-exascale and exascale supercomputers.[65] Deployment of AI Factories began with selections in late 2024, when seven initial consortia were awarded to host facilities, followed by expansions in 2025.[66] On October 10, 2025, EuroHPC JU selected six additional sites in the Czech Republic, Lithuania, the Netherlands, Romania, Spain, and Poland, bringing the total toward a network of 19 factories.[14] [63] To enhance connectivity, 13 AI Factory Antennas—regional extensions linking to core factories—were designated on October 13, 2025, including connections in Poland and Moldova, with an EU investment of approximately €55 million matched by national contributions.[16] [67] Earlier phases incorporated sites in Czechia, Denmark, Estonia, Norway, Poland, Portugal, Romania, and Turkey, with combined investments reaching €485 million from EU and national sources for select deployments.[64] [68] Access prioritizes European data and models, with policies ensuring non-discriminatory use while restricting transfer to foreign entities, though critics note potential scalability limitations compared to private-sector hyperscalers.[69] The InvestAI Initiative complements AI Factories by scaling infrastructure through larger "gigafactories," launched by the European Commission on February 10, 2025, to mobilize €200 billion in public-private investments for advanced AI development.[70] Allocating €20 billion specifically for up to five gigafactories—each equipped with around 100,000 latest-generation AI chips, approximately four times the capacity of standard AI Factories—the program targets training of highly complex, large-scale models while maintaining European control over data and algorithms.[70] [71] This builds directly on EuroHPC JU's framework, repurposing and augmenting its supercomputing assets to address Europe's lag in proprietary AI hardware, with public funding mechanisms involving the EuroHPC JU for site selections and operations.[72] Gigafactories represent the world's largest planned public-private AI partnerships, focusing on trustworthy systems, though their success depends on attracting private matching funds amid competition from U.S. and Chinese investments.[70] [73]DARE Data Centre Project
The DARE project, formally known as Digital Autonomy with RISC-V in Europe (Specific Grant Agreement 1, or SGA1), represents a key initiative under the EuroHPC Joint Undertaking to foster European technological independence in high-performance computing (HPC) and artificial intelligence (AI) through open-source RISC-V architecture. Launched on March 6, 2025, it addresses Europe's reliance on non-European semiconductor technologies by developing indigenous hardware and software stacks tailored for exascale supercomputing and AI workloads.[28][74] The project aligns with broader EU goals of digital sovereignty, emphasizing energy-efficient, customizable processors to compete with proprietary architectures from global leaders.[75] Coordinated by the Barcelona Supercomputing Center (BSC), DARE involves a consortium of 38 partners from academia, industry, and research institutions across Europe, including entities like Forschungszentrum Jülich, imec, and Codasip. Funded with €240 million from the EuroHPC JU under grant agreement No. 101202459, the three-year first phase focuses on prototyping a complete European supercomputing ecosystem using advanced silicon nodes.[76][77] Key technical objectives include designing three specialized RISC-V-based chiplets: a high-end general-purpose scalar processor for core computing tasks, a vector accelerator for high-throughput numerical simulations, and an AI inference accelerator optimized for low-power edge-to-cloud deployments.[77][78] These components aim to enable scalable, modular systems capable of supporting critical applications in domains such as climate modeling, drug discovery, and materials science, with input from leading HPC centers to ensure compatibility with real-world demands.[28][79] While primarily hardware-focused, DARE incorporates software ecosystem development, including open-source compilers, runtime environments, and middleware to integrate the chiplets into full HPC prototypes deployable in European data centers. The initiative builds on RISC-V's extensible instruction set to prioritize performance-per-watt metrics, targeting reductions in energy consumption for large-scale data center operations compared to x86 or ARM alternatives.[74][80] Partners like Axelera AI contribute to the AI chiplet, securing up to €61.6 million specifically for scalable inference hardware that enhances data center efficiency in AI factories.[81] This phase sets the foundation for subsequent stages, potentially culminating in production systems like an upgraded MareNostrum 6 supercomputer, with demonstrations planned to validate interoperability and sovereignty in data center environments.[82][83] EuroHPC JU's selection of DARE underscores a strategic pivot toward "designed-in-Europe" technologies, mitigating supply chain vulnerabilities exposed by global chip shortages and geopolitical tensions. Evaluations by consortium experts emphasize empirical benchmarks for throughput, latency, and fault tolerance, rather than unverified vendor claims, to ensure viability for data-intensive workloads in sovereign infrastructures.[28][84] The project's progress is monitored through milestones tied to prototype integration and testing in simulated data center setups, with public reporting via the EuroHPC JU framework to promote transparency and adoption across member states.[85]Research and Innovation Funding Calls
The EuroHPC Joint Undertaking issues Research and Innovation (R&I) funding calls to advance high-performance computing capabilities across Europe, focusing on technology development, application enablement, skills enhancement, and international collaboration. These calls, primarily funded through the Horizon Europe programme, follow open and competitive procedures outlined in the JU's annual work programmes, targeting priorities such as energy-efficient HPC systems, quantum integration, AI-driven applications, and support for small and medium-sized enterprises (SMEs) to adopt HPC tools. Grants support consortia of public and private entities to build a sovereign European HPC ecosystem, reducing reliance on non-EU technologies.[86][87] Specific calls address targeted gaps in the HPC value chain. For instance, on November 8, 2023, the JU launched three R&I calls, including one for Energy Efficient Technologies in HPC to optimize power consumption in supercomputing infrastructures and another for the EuroHPC Virtual Training platform to standardize skills development across member states.[88] In early 2025, a call opened on March 4 to strengthen international cooperation on artificial intelligence, with submissions due by June 4, 2025, aiming to integrate global AI advancements with European HPC resources.[89] By July 1, 2025, two additional calls were announced to bolster HPC skills and support, emphasizing diversity in participation and global visibility for European expertise.[90] Recent initiatives include the Quantum Grand Challenge call launched on October 14, 2025, which provides Horizon Europe grants in two phases to European quantum computing startups developing market-ready solutions compatible with EuroHPC systems.[20] [91] Broader efforts, such as the €60 million allocation for High-Performance Computing Centres of Excellence, fund software and algorithm optimization to sustain Europe's competitiveness in exascale computing.[92] Projects like FFplus, supported through cascading open calls, facilitate SME uptake of HPC and AI by funding sub-projects for process optimization and innovation.[93] These calls have resulted in 57 funded R&I projects as of 2025, yielding tangible advancements such as the SEANERGYS initiative for energy-efficient software on EuroHPC supercomputers, the DARE project for RISC-V-based hardware and software co-design toward exascale systems, and EVITA for a unified HPC training framework.[94] Outcomes include enhanced supply chain resilience, with evaluations prioritizing measurable impacts on computational performance, energy savings, and industrial applications over unsubstantiated projections.[95]Membership Composition
Participating States and Observers
The EuroHPC Joint Undertaking consists of 37 participating states as of October 2025, encompassing all 27 European Union member states along with 10 associated third countries that have acceded to the framework.[1] These states provide matching funds, host infrastructure, and appoint representatives to the Governing Board, which oversees strategic decisions and resource allocation.[96] Membership enables coordinated investments in high-performance computing, artificial intelligence, and quantum technologies, with contributions scaled to each state's economic capacity under the Joint Undertaking's statutes.[97] The full list of participating states, in alphabetical order, is as follows:- Albania
- Austria
- Belgium
- Bulgaria
- Croatia
- Cyprus
- Czechia
- Denmark
- Estonia
- Finland
- France
- Germany
- Greece
- Hungary
- Iceland
- Ireland
- Israel
- Italy
- Latvia
- Lithuania
- Luxembourg
- Malta
- Moldova
- Montenegro
- Netherlands
- North Macedonia
- Norway
- Poland
- Portugal
- Romania
- Serbia
- Slovakia
- Slovenia
- Spain
- Sweden
- Türkiye
- United Kingdom[1][98]