Italy Enters the AI Infrastructure Investment Map

Italy has entered the map of major European investments in infrastructure for artificial intelligence, and among the operators who have decided to bet on the country is Khazna Data Centers, a company specializing in hyperscale data centers—massive facilities designed for the most intensive cloud and AI workloads, based in the United Arab Emirates. The project surrounding its entry into Southern Europe is a 500 MW campus that the company is developing together with Eni in Ferrera Erbognone, Lombardy, one of the largest AI infrastructure plans announced on the continent.

Greg Jasmin, Chief Commercial Officer of Khazna, who has over 27 years of experience in digital infrastructure, illustrates the business strategy behind this choice. Khazna develops and manages large data centers for clients needing significant capacity, high reliability, and rigorous execution. Jasmin summarizes the model in three pillars: rapid hyperscale realization, large-scale operational excellence, and infrastructure designed for the AI era. The difference from a traditional operator, he claims, lies not in the individual data center but in how it is designed and managed. The demand linked to AI brings higher densities, different cooling requirements, and tighter expectations on testing and service continuity. Hence, the choice to standardize designs to make them repeatable and to size power and cooling in anticipation of future densities.

The key point that Jasmin emphasizes is the overcoming of a separate-phase approach.

"Traditional data center development can treat power, cooling, and construction as sequential problems. We treat them as an integrated system because, at the scale of AI, it is the integration that protects timing, reliability, and the overall long-term cost."

This position serves as a premise for everything else: the commitment to deliver capacity in predictable timelines holds only if the three elements are engineered together from the start.

Why Italy

The interest in the Italian market arises, according to Jasmin, from rapidly growing demand and clients who want high-performance capacity closer to users and business operations in Southern Europe. The local fundamentals also weigh heavily: network connectivity, resilient power, and the presence of a partner capable of building large infrastructure while respecting the real constraints of a construction site.

From a commercial standpoint, the approach is explicitly selective. Khazna prioritizes hyperscalers, the major cloud and AI service providers that purchase capacity on a vast scale, and large enterprises seeking long-term collaborations and clear growth paths, focusing on campus development that can be realized in phases.

"We build campuses that can be delivered in phases, so clients can expand as demand grows without having to start over each time from site selection and the authorization process," explains Jasmin. In parallel, the company says it will soon engage with customers on the requirements that really matter for AI workloads, from density ranges to cooling preferences to operational readiness, so that what is built corresponds to actual workloads and not generic capacity assumptions.

The Campus with Eni and the Path to 1 GW

The 500 MW campus aims, according to Jasmin, to create large-scale digital infrastructure and infrastructure that supports the next wave of AI and cloud demand in Italy and the surrounding area, with energy considerations integrated from the outset thanks to the partnership with Eni.

"At this scale, the distinguishing factor is not building a building, but delivering repeatable blocks of capacity with predictable lead times, reliability, and operational maturity," he observes. The campus is primarily designed for large clients needing significant capacity and the ability to grow in phases, typically hyperscalers and sovereign AI projects, the national infrastructure that various states want to maintain control over their data.

Moreover, the campus is just the first step in a broader plan. The letter of intent signed by the United Arab Emirates and Italy aims to achieve a total capacity of up to 1 GW in the country, with the 500 MW of Ferrera Erbognone representing the first milestone. On this horizon, Jasmin maintains a cautious profile and prefers to be precise about the process: after signing the preliminary agreement (Head of Terms) and finalization, subject to the approval of the competent authorities on the joint venture, the priority becomes execution. Again, the method is the same; engage with clients early to align the campus development plan with actual demand profiles, especially regarding density, cooling preferences, and operational requirements.

Then there’s the issue of the technologies with which the site will support AI workloads, from direct liquid cooling to energy efficiency measured with PUE (Power Usage Effectiveness), to power density per rack. In this area, Jasmin deliberately chooses not to anticipate numbers. An infrastructure ready for AI, he explains, arises from designing for a range of densities and being able to accommodate next-generation cooling approaches as customer needs evolve, including liquid cooling where necessary. More than the single technological choice, he asserts, it counts that power and cooling are engineered as a single system, rigorously tested, and managed on measured performance standards.

Regarding PUE values and density, in particular, he prefers not to set targets before the project is defined.

"We prefer not to publish specific targets before the projects are finalized. They depend on design choices, climatic conditions, and especially customer load profiles and cooling preferences."

The same applies to timing: while recognizing that the market would like precise dates, he reiterates that the actual operational phase will only commence after the preliminary agreement, finalization, and clearance from the authorities. A caution that, more than evasive, reflects an industrial culture—one that prefers not to announce numbers that the final project might later contradict.

The Energy Component and Model Replicability

The energy component is what makes the Italian project unique. The campus integrates computing capacity with Eni's Blue Power, low-emission electricity, and CO2 capture through the Ravenna CCS project. Jasmin, however, invites one not to read it as a simple combination of computing and a single decarbonization mechanism, preferring to speak of an operational and measurable ESG approach, characterized by design efficiency, choice of cooling suitable for the environment and density, construction standards that reduce waste and rework, and transparent performance monitoring.

The question that matters most to the market is whether this model is replicable elsewhere or remains linked to the Italian industrial context. For Jasmin, the lesson that can be exported is not the single facility but the principle, that is, to integrate the data center's realization with a long-term energy strategy.

"The most lasting advantage comes when the data center operator and the energy partner plan together: capacity phases, interface with the grid, resilience, and the path toward lower-emission energy over time," he states. Italy has its specifics, but he argues that the logic of partnership can travel: wherever the availability of energy and sustainability requirements are the true factors that condition the realization of infrastructure for AI, closer alignment between infrastructure and energy planning becomes a competitive necessity, not merely an option among many.

It is on this ground that Khazna attempts to distinguish itself from traditional operators and position Italy as a gateway to low-latency AI services for Southern Europe. The test bed of Ferrera Erbognone will not be so much the installed power, but rather the capacity to keep energy, cooling, and computing together in a single industrial design, respecting timelines that the AI market does not forgive.