EuroHPC JU has signed a contract with E4 Computer Engineering and Dell Technologies to deploy a new AI-optimized supercomputer at IT4LIA, the Italian AI factory hosted and operated by CINECA at the DAMA Tecnopolo in Bologna. The system is part of the European strategy to open supercomputing capacity to companies, startups, small and medium-sized enterprises, and research centers that cannot build infrastructure of this scale on their own.
The planned investment for the acquisition, delivery, installation, and maintenance of the system amounts to €290 million, co-funded 50% by EuroHPC JU through the Digital Europe program and by Italy’s Ministry of University and Research. This figure is significant because it confirms that the European Union’s commitment to AI goes beyond regulation, talent, or language models: it is also investing in physical infrastructure to train, tune, and run large-scale AI workloads.
A system based on NVIDIA Blackwell and liquid cooling
The new supercomputer will be manufactured by Dell Technologies and integrated by E4 Computer Engineering. The chosen architecture combines NVIDIA GB200 NVL4 systems with liquid cooling, NVIDIA Grace processors, NVIDIA Blackwell GPUs, and NVIDIA Quantum-X800 InfiniBand networking. According to EuroHPC JU, the system is designed to deliver over 160 exaflops of peak performance for AI inference.
This figure should be interpreted carefully. It does not directly equate to traditional double-precision floating-point performance used in classical scientific supercomputing rankings. Instead, it refers to peak performance for AI inference—a metric linked to lower-precision operations and workloads typical of advanced models. Still, this represents a significant leap for Europe, because inference has become a critical part of the actual cost of AI.
The system will also incorporate VAST Data’s data platform, designed to provide a unified data layer for AI workloads. This element is not secondary. In AI infrastructure, moving, preparing, serving, and protecting data can be as complex as deploying accelerators. Without a fast, coherent data architecture, GPUs may spend time waiting for data rather than operating at full capacity.
EuroHPC JU also highlighted a dedicated inference partition with accelerators from Axelera AI, a European company specializing in AI chips, and CPUs designed by SiPearl. This combination aims to reinforce a political and technological narrative: Europe needs access to global accelerators like NVIDIA, but also wants to incorporate European components into the AI value chain.
IT4LIA, an AI factory for industry and research
IT4LIA does not start from scratch. Italy’s AI factory already leverages the supercomputer Leonardo, its upgrade LISA, the GAIA cloud system, MEGARIDE, and the new AI-optimized system planned for Bologna. Except for MEGARIDE, located at the San Giovanni a Teduccio campus in Naples, all systems are housed at the DAMA Data Manifattura Tecnopolo in Emilia-Romagna, Bologna. Overall, the infrastructure boasts over 20,000 GPUs for AI workloads, from data preparation to training and inference.
The goal of a European AI Factory is not just to showcase power. Its purpose is to provide computing capacity, technical services, data access, training, and support to companies and research centers. In the case of IT4LIA, EuroHPC JU highlights sectors such as agri-food, cybersecurity, meteorology and climate, and manufacturing. Horizontal services for secure data management, analysis, metadata, and compliance verification with Italian and European regulations on data and AI also are planned.
This aspect is especially important for SMEs and startups. Many European companies possess sector-specific knowledge, proprietary data, and valuable use cases but lack direct access to large GPU clusters or specialized HPC equipment. The network of AI factories aims to bridge this gap: reducing the divide between research, public infrastructure, and enterprise adoption.
EuroHPC JU states that IT4LIA is primarily designed for startups and SMEs, though it will also be accessible to the research community. Since April 2025, the Italian factory has already offered AI computing resources and services based on existing infrastructure, mainly the Leonardo supercomputer. The new system will expand this capacity and add a layer more oriented toward modern AI workloads.
A piece of Europe’s technological sovereignty
The signing of this contract occurs amid a global race for AI infrastructure. The United States boasts large hyperscale clouds and a supply chain dominated by NVIDIA, AMD, Broadcom, Marvell, and others. China invests to reduce reliance on export-controlled technologies. Europe, with fewer major cloud providers and a smaller industrial base in accelerators, seeks to build a pathway supported by public supercomputing, research centers, specialized tech companies, and data sovereignty norms.
The EuroHPC AI Factories network is part of that strategy. According to EuroHPC JU, the European Union has established 19 AI factories and 13 associated hubs to provide free and customized support to SMEs and startups. IT4LIA is one of Italy’s nodes within this map, centered in Bologna—a hub already vibrant with supercomputing, big data, AI, and quantum technologies.
The role of E4 Computer Engineering also has an industrial significance. The Italian company will be responsible for system integration and has highlighted that this contract makes it the first Italian integrator to win a tender within the European AI Factories initiative. Their statement mentions over 2,000 nodes, 8,000 GPUs, as well as software stack development, support, and maintenance in production.
Similarly, Dell Technologies’ involvement highlights the pragmatic limits of Europe’s strategy. While aiming for greater autonomy, Europe’s major AI systems still largely depend on global suppliers for servers, GPUs, networking, and software. The key difference lies in who operates the infrastructure, under what access conditions, with which data protection standards, and how much of the European ecosystem participates in its design, integration, and usage.
The IT4LIA project will not single-handedly close Europe’s gap with the US or China. Nor does it replace the need for more private investment, better models, talent, energy, and real-world application deployment. However, it provides a concrete piece: accessible, public computational capacity for European companies and researchers to test, train, and deploy applications without always relying on foreign infrastructure.
The real key will be in how it’s used. An AI supercomputer is only valuable if it reaches those capable of transforming it into applied innovation—labs, startups, manufacturers, hospitals, government agencies, universities, or industry. IT4LIA aspires to be just that—a gateway to advanced AI for organizations needing compute power, technical support, and a framework aligned with European privacy and sovereignty standards.
Frequently Asked Questions
What has EuroHPC JU signed with Dell and E4?
EuroHPC JU has signed a contract with E4 Computer Engineering and Dell Technologies to deploy an AI-optimized supercomputer at the Italian factory IT4LIA, operated by CINECA in Bologna.
What is the cost of the new IT4LIA supercomputer?
The budget for acquiring, delivering, installing, and maintaining the system is €290 million, co-funded by EuroHPC JU and Italy’s Ministry of University and Research.
What technology will the system use?
The supercomputer will be based on NVIDIA GB200 NVL4 architecture with liquid cooling, Blackwell GPUs, Grace CPUs, Quantum-X800 InfiniBand networking, VAST Data’s data platform, and a dedicated inference partition featuring European technology from Axelera AI and SiPearl.
Who will be able to access IT4LIA?
The system is mainly targeted at startups, SMEs, and the research community, with services supporting sectors such as agri-food, cybersecurity, meteorology, climate, and manufacturing.