Vertiv Expands Croatian Capacity for AI Data Centers

The growth of artificial intelligence is no longer only about new models, applications and computing chips. Behind them, the physical infrastructure required to run AI workloads is changing just as quickly, with data centres becoming more power-intensive, more complex and increasingly sensitive to deployment speed. Demand that was particularly visible in the United States over the past two years is now accelerating across Europe, prompting Vertiv to expand its engineering and manufacturing capacity in Croatia.

“Demand for data centres in Europe has really exploded this year,” Vedran Brzić, Vice President of Infrastructure Solutions for EMEA at Vertiv, tells ICTbusiness Media – ICTbusiness.info & ICTbusiness TV. What the industry saw in the US during the previous two years is now increasingly visible in Europe, he explains. That growth directly increases the need for both manufacturing capacity and engineering expertise capable of designing, integrating and delivering increasingly complex data-centre infrastructure.

Vertiv is consequently opening more than 200 skilled positions in Croatia. The company already employs more than 650 people in Zagreb and Rugvica, supporting large-scale data-centre projects. More than 50 new engineering positions are planned across electrical, mechanical and civil engineering, alongside more than 155 roles for electricians, assembly specialists and cooling technicians.

Brzić stresses that this is not simply an expansion of manufacturing. “We are also increasing the size of our engineering teams – electrical, mechanical and civil engineers, automation specialists and architects,” he says. AI infrastructure is not merely about installing more servers. It changes power requirements, cooling architecture, project design and, increasingly, the amount of time customers are prepared to wait before new capacity becomes operational.

One of the most important changes is taking place in the construction model itself. Instead of assembling every complex infrastructure system only at the final data-centre site, more integration is moving into a controlled factory environment. “The idea is to integrate the more complex infrastructure systems in the factory. The installation process is cleaner, faster and more efficient,” Brzić explains.

The benefits become more pronounced as projects increase in scale. When deployments involve a series of 50 or 100 similar modules rather than a handful of individual systems, factory integration can make the process considerably more efficient and repeatable. Modules may contain power, cooling or IT infrastructure and can be tested before leaving the manufacturing facility. “Once that block arrives at the final site, commissioning time is very short,” Brzić says.

That matters in a market where customers want to put new computing capacity into commercial operation as quickly as possible. Data centres represent major capital investments, making the period between project launch and revenue-generating operation an increasingly important part of the business case. Vertiv is therefore overlapping stages that would traditionally have been completed sequentially.

“While the design is still in progress, we are already ordering equipment and components and starting integration,” Brzić says. Running design, procurement and manufacturing in parallel can shorten the overall project schedule. Rather than waiting for every design phase to be completed before production starts, selected activities can proceed simultaneously so that modules are ready when the site is able to receive them.

Alongside speed, power density is becoming the other major challenge. AI workloads concentrate increasing amounts of computing power into relatively limited physical space, directly raising the requirements for electrical and thermal infrastructure. “As power density increases, power infrastructure and cooling infrastructure take up an increasingly important part of the overall data-centre design,” Brzić explains. The space housing the IT equipment itself is no longer the only dominant element of a project.

Power delivery and heat removal are therefore becoming fundamental design constraints, while infrastructure also has to remain sufficiently flexible to accommodate future expansion. “We very often take potential future expansion into account,” Brzić says. A modular approach makes it possible to add new power blocks and other infrastructure as demand increases rather than redesigning the entire facility whenever additional capacity is required.

The pace of hardware development makes that adaptability increasingly valuable. Brzić points to rapidly changing technology generations, each bringing its own infrastructure requirements. Some developments can be anticipated from technology road maps published by manufacturers, but data-centre architecture still needs enough flexibility to absorb changes that cannot be predicted precisely at the start of a long-term infrastructure project.

Brzić expects some of the most significant changes to emerge in the power architecture itself. Very large amounts of electrical power increasingly have to be delivered into comparatively small spaces, affecting distribution systems inside the facility as well as the relationship with the wider grid. He also points to the possibility of combining grid power with generation located directly at the site, including solar systems or gas generators, depending on the project and its requirements.

For Vertiv, the expansion in Croatia is therefore directly linked to the broader transformation of the data-centre market. Teams in Zagreb and Rugvica contribute to the design, development and delivery of integrated power and thermal-management systems intended for high-density data centres and increasingly demanding AI workloads.

Croatia’s role in that model extends beyond manufacturing alone. It brings engineering, project design, integration and production together within the same regional delivery structure. “We definitely have to keep pace with the growth we expect to see this year and in the years ahead,” Brzić concludes. The direction is already clear: as AI computing becomes denser and more powerful, data-centre deployment increasingly resembles an industrialised infrastructure process in which modularity, power, cooling and deployment speed determine how quickly new capacity can begin operating.