Skip to main contentSkip to footer
Logo IT4LIA AI factory

HPC technological infrastructure

One of the most powerful infrastructure in Europe

IT4LIA AI Factory benefits from the EuroHPC LEONARDO supercomputer, its LISA upgrade, the GAIA cloud system, MEGARIDE, and the new IT4LIA supercomputer, optimised for artificial intelligence, whose development contract was signed in April 2026. All systems, with the exception of MEGARIDE located at the San Giovanni a Teduccio University Campus in Naples, are hosted in the data center of DAMA Data Manifattura Tecnopolo Emilia-Romagna in Bologna. Together, they provide a complete computing continuum that supports the rapidly growing AI ecosystem in Italy and across Europe.

By combining these systems, the technological infrastructure delivers over 20,000 GPUs to power Artificial Intelligence workloads, from data preparation to model training and inference.

The total number of GPUs is set to grow significantly with the entry into operation of the new IT4LIA supercomputer.

The entire IT4LIA AI Factory infrastructure is integrated into the European supercomputing network and operated by CINECA.

LEONARDO
LISA
GAIA
MEGARIDE
IT4LIA

Leonardo

  • 155 system racks
  • 4,992 compute nodes
  • 250 Petaflops
  • 2,800 TB RAM
  • 6 MW power usage
  • 110 PB storage
  • 600 sqm footprint
  • More than 95% heat dissipation via DLC

LEONARDO is a pre-exascale supercomputer with a peak performance of nearly 250 PetaFLOPS and over 100 PB of storage capacity. It represents a major step forward in providing pre-exascale computing capabilities to Italian and European researchers.

With 3,500 processors and 14,000 state-of-the-art GPUs, it ranks among the world’s top ten most powerful supercomputers. The system integrates advanced processing components to handle highly complex computational workflows that combine HPC, AI, high-throughput computing and visualization applications.

For more details, please refers to the article at the following link.

Blue server cabinets labeled LEONARDO showcase advanced HPC technological infrastructure in a modern, brightly-lit data center.
Rows of large black HPC server cabinets labeled Leonardo in a modern data center with European Union logos displayed.

Lisa

LISA is the new AI-optimized partition within the LEONARDO system. It targets a sustained double-precision peak of 65 PFlops and a mixed-precision HPL-MXP61 performance of 650 PFlops.

The system uses 8-way GPU nodes with at least 640 GB of high-bandwidth shared memory, enabling the training of some of the largest AI models. LISA is the first EuroHPC compute partition designed from scratch specifically for AI workloads.

LEONARDO and LISA are connected through the same multi-protocol storage tier, which provides simultaneous block, file, and object access. This configuration enables the essential data services required to support AI computing tasks.

Gaia

  • 420 interactive CPU nodes
  • 80 interactive GPU nodes
  • 10 PB storage

GAIA complements and supports AI and HPC services through cloud technologies. The system integrates a large-scale data lake that supports multiple protocols and is entirely based on full-flash technology, ensuring high performance and scalability for diverse workloads.

It is designed to help data scientists maximize productivity in data collection, cleansing, and transformation, in implementing and testing advanced machine learning algorithms, and in leveraging Big Data analytics for business insights.

GAIA is a ready-to-go appliance that combines the power of a multi-GPU server with a fully containerized software stack. It includes tools for data management, interactive computing environments based on Notebook technology, services for visualizing trained models, and automation tools for data science workflows.

Together with the EuroHPC infrastructure, GAIA forms a compute continuum that can respond flexibly and in an integrated manner to the different needs of the ecosystem.

Megaride

  • 340 GPUs
  • 28,000 high-performance computing cores
  • 50 PB of ultra-fast storage
  • Liquid-cooling technologies for sustainable and efficient operation

Located at the San Giovanni a Teduccio University Campus in Naples, MEGARIDE takes its name from the small islet of Naples where Castel dell’Ovo stands. It is one of the most energy-efficient supercomputing systems in Europe, designed to handle the full range of AI use cases.

From model development and adaptation through the AI/HPC (scale-out) partition to the definition and hosting of new services on the OpenStack cloud partition, MEGARIDE offers cutting-edge computing capabilities. The cloud partition uses the same data repository and virtualization technologies as GAIA, creating a geo-distributed cloud and data lake system.

The infrastructure was developed through an initiative by the Ministry of Universities and Research and the National Cybersecurity Agency, with the support of ICSC – National Research Centre in HPC, Big Data and Quantum Computing. MEGARIDE is a state-of-the-art facility built to meet the computing, security, and sustainability needs of the National Cybersecurity Agency, serving as a dedicated cybersecurity hub that leverages AI and machine learning to protect digital infrastructures and drive technological innovation.

IT4LIA

  • Peak AI inference performance: >160 Exaflops
  • Architecture: NVIDIA GB200 NVL4 (liquid-cooled)
  • CPU: NVIDIA Grace + SiPearl
  • GPU: NVIDIA Blackwell
  • Network: NVIDIA Quantum-X800 InfiniBand
  • Storage: VAST Data platform
  • Budget: 290 million euros
  • Location: DAMA Tecnopolo, Bologna

The new IT4LIA supercomputer represents the next generation of the IT4LIA AI Factory infrastructure. The contract for its development was signed on 22 April 2026 by EuroHPC JU with E4 Computer Engineering and Dell Technologies. The system will be installed at the DAMA Tecnopolo Data Manifattura Emilia-Romagna in Bologna and managed by CINECA.

With a peak performance of over 160 Exaflops for AI inference, it will rank among the most advanced infrastructures in Europe. The liquid-cooled NVIDIA GB200 NVL4 architecture combines NVIDIA Grace CPUs with NVIDIA Blackwell GPUs and an NVIDIA Quantum-X800 InfiniBand network, with the VAST Data platform providing a unified data layer for all AI workloads. A dedicated inference partition will include AI-optimised accelerators from Axelera AI alongside SiPearl CPUs — both of European design — reinforcing the continent’s technological sovereignty.

The system is primarily intended for startups and SMEs, and is also open to the research community, in full compliance with European data sovereignty and privacy standards.

For further information, please refer to the following article: link.

Why HPC is important for Artificial Intelligence

Artificial Intelligence requires massive amounts of data and computing power to reach its full potential. This is where the IT4LIA AI Factory infrastructure makes the difference, because High Performance Computing (HPC) can process in seconds what a traditional system would take days or even weeks to complete.

In AI, speed is not a luxury but a competitive advantage. Training complex machine learning and deep learning models means working on huge datasets and constantly testing new architectures and parameters. With HPC, these training cycles become much shorter, delivering more accurate results in less time and enabling rapid iterations to improve algorithm performance.

With high-performance GPUs, ultra-fast networks, and massive parallel processing capacity, an HPC infrastructure can:

  • Analyze petabytes of data in real time, extract hidden patterns and insights, and power predictive AI systems.
  • Train large AI models on massive datasets within competitive timeframes
  • Accelerate AI deployment into production, reducing costs and shortening time-to-market.

For IT4LIA AI Factory, HPC is not simply a technological support. It is the engine that powers more powerful, precise, and scalable AI, capable of addressing challenges in fields ranging from healthcare to finance, from advanced manufacturing to security. In a world where data is growing exponentially and decisions must be taken faster than ever, supercomputing turns AI from a promise into an operational reality.