Colosseum supports the deployment of enterprise-grade sovereign AI infrastructures. Combines exceptional processing power, energy efficiency and data sovereignty, while balancing the most stringent data security requirements
Vertiv, (NYSE: VRT), a global provider of critical digital infrastructure and continuity solutions, has partnered with NVIDIA and AI leader iGenius to implement Colosseum, one of the world's largest NVIDIA DGX AI supercomputers, powered by NVIDIA Grace Blackwell Superchips. Expected in Italy by the end of 2025, Colosseum will redefine the digital environment thanks to a unique AI Sovereign data center for regulated workloads.
Colosseum meets the needs of highly regulated sectors such as finance, healthcare and public administration and will represent the union of advanced computing power, energy efficiency and data sovereignty, while complying with the most stringent data security requirements.
Colosseum, an NVIDIA DGX SuperPOD, is the latest addition to a multi-year collaboration between Vertiv and NVIDIA. It is strategically positioned in southern Italy to meet the needs of regional administrations, marking a milestone in the context of AI in Europe.
“Leveraging the power of NVIDIA’s cutting-edge accelerated processing and Vertiv’s innovative infrastructure expertise, Colosseum demonstrates the evolutionary potential of Sovereign AI,” said Uljan Sharka, CEO of iGenius. “It demonstrates how modular systems and software interfaces enable a new era of mission-critical AI.”
The design is modular and is designed for efficiency
Colosseum combines Vertiv's infrastructure building expertise, NVIDIA's accelerated computing, and NVIDIA's Omniverse Blueprint for designing and operating AI data centers. Deployment will leverage Vertiv 360AI data center power and cooling infrastructure platform, designed for NVIDIA GB200 NVL72, developed in collaboration with NVIDIA and released at the end of 2024.
This modular and scalable system allows iGenius to deploy one of the fastest hyperscale AI supercomputers and one of the largest capable of supporting AI Sovereign.
Vertiv has also expanded its own reference design library AI Hub with co-developed design for powering and cooling data centers for NVIDIA GB300 NVL72. Staying one step ahead of the GPU generation, Vertiv allows customers to plan infrastructure before silicon arrives, with ready-made designs that feature higher rack power density and replicable models for large-scale AI factories.
“The computing unit is no longer the single chip, it is the system, the AI Factory,” underlined Karsten Winther, president of Vertiv in EMEA. “Through collaborations with NVIDIA and iGenius, an AI visionary, we are demonstrating system-level efficiency and maturity in offering a data center as the compute unit, driving the rapid adoption of AI-native power and cooling infrastructure, which serve as a catalyst for large-scale AI adoption.”
Simulation with NVIDIA Omniverse. Delivery with speed
“AI is redefining the data center landscape, requiring new standards of scale, efficiency and adaptability for global AI factories,” said Charlie Boyle, vice president of DGX platform at NVIDIA. “With physically structured digital twins enabled by NVIDIA Omniverse technologies and Vertiv's modular design for the iGenius DGX SuperPOD data center, Colosseum sets a new standard for building supercomputers for the AI era.”
Colosseum is co-designed as a physically precise digital twin, developed using technologies NVIDIA Omniverse, which enables real-time collaboration between Vertiv, iGenius and NVIDIA, accelerating system-level decisions and reducing the design-to-deploy cycle. Omniverse Blueprint offers real-time simulations, allowing engineers to test and refine designs instantly, reducing simulation process times from months to hours. The production and integration phases in the Vertiv factory reduce implementation times by up to 50% compared to the construction of traditional data centers.
The collaborative 3D design validated the entire infrastructure stack, enabling predictive modeling of thermal load, electrical flow and site layout, from 132 kW liquid-cooled racks to modular power systems, before each individual module was even built.
Intelligently designed. Optimized by software
Vertiv's prefabricated, AI-ready modular data center solution is designed, built, delivered, installed and tested by Vertiv. It includes solutions for power, cooling, management, monitoring, support and maintenance, with power and cooling capacity initially supporting up to 132kW/rack, with the ability to scale to future needs. The architecture of the building integrates ready-to-use white space areas inside and those for complete modular gray space outside. This approach offers exceptional scalability and energy efficiency, revolutionizing the way data centers are built and deployed.
Colosseum will leverage NVIDIA Mission Control for data center operations and orchestration, and Vertiv™ Unify will be able to simplify and synchronize building management for AI factories. Vertiv Unify offers:
- Orchestration of power, cooling and computing services in real-time
- Digital twin synchronization for closed-loop optimization
- AI-ready capabilities that support autonomous decision making
Thanks to the integration of NVIDIA Omniverse technologies, Vertiv Unify is able to update data from physical systems and digital models in real-time, enabling predictive maintenance, "what-if" simulations and verification of environments before operational risks arise.
Blueprint for the creation of AI Factories globally
Colosseum is much more than a data center. It is the prototype for scalable, replicable, Sovereign-like AI structures. Combining cloud-level density, local data control and modular deployment, it signals the future of AI where inference must be certain, fast, compliant and distributed.
This is not an isolated project, but it is a point of reference. iGenius creates a model with Colosseum that is destined to be replicated globally, with Vertiv and NVIDIA in perfect harmony on the support of future platforms, including DGX GB300 systems and beyond. The future of AI Sovereign is no longer theoretical: it is underway.
