At the LCL Wallonia One Event, Vladimir Prodanović, Principal Program Manager at NVIDIA, delivered a standout presentation that captivated the audience. His keynote painted a clear picture of where AI infrastructure is heading and how quickly the landscape is evolving. Prodanović elaborated on these trends in an interview that followed his keynote.
The shift from traditional IT loads to extreme AI workloads is reshaping data center design as we know it. Rack density is no longer measured in kilowatts but in hundreds of kilowatts, moving rapidly towards the megawatt era.
This evolution places new demands not only on power and cooling, but also on rack handling and installation. AI infrastructure redefines what racks, rooms and entire buildings must support. According to Prodanović, new racks support up to 1MW and can weigh nearly 4 tonnes. Upcoming systems will be even heavier. This shift forces operators to question common assumptions about raised floors, logistics, floor loading and the design of goods lifts and access routes.
Cooling requirements shift as well. As rack densities approach the megawatt range, direct liquid cooling becomes the standard, with air cooling handling the remaining heat. These requirements make modular building design essential, allowing power and liquid connections to be reconfigured quickly and safely. As Prodanović explains, 1MW is only a milestone.
One of the most forward-looking developments inside NVIDIA is the use of digital twins to design data centers. Prodanović describes how before any concrete is poured or racks are installed, entire sites are now simulated and optimised in software. As Prodanović puts it: “We are using AI to create AI. Clusters are built first in software, adjusted based on testing, and then deployed in the real world. Once operational data is available, it is reintroduced into the simulation to refine the model. This approach supports the development of facilities ranging from relatively modest AI clusters to gigawatt-scale campuses.”
Digital twins allow NVIDIA and its partners to test layouts, cooling patterns, power distribution and scaling scenarios. This way of working also enables closer collaboration, as co-design becomes far more practical when all stakeholders can evaluate options inside the same virtual environment. For operators, it offers a way to shorten project timelines and reduce uncertainty, especially as AI infrastructure evolves at a rapid pace.
Although hyperscalers dominate the upper end of the market, Prodanović is clear about the continued importance of local and regional data centers. Sovereignty, flexibility but also community are critical for building a strong local AI ecosystem.
He highlights the strong collaborative culture in countries like Denmark, where companies, universities and data centers co-invest in national AI infrastructure. Their sovereign cloud initiatives benefit from a tradition of local ownership and smaller, regionally operated infrastructure. In other regions, where dependence on hyperscale cloud has grown over many years, this model of collaboration is more difficult.
Prodanović emphasises that success in AI is not limited to the largest players. The real transformation will come when more companies know how to use these technologies in their daily operations. Local data centers enable that shift by offering edge locations that bring compute closer to end users.
NVIDIA’s internal culture is a core part of its competitive edge. With around 39,000 employees, the company operates with the agility of a start-up and the technical depth of a research lab. Everyone stays close to customers and works at the pace of the industry’s evolution.
Prodanović illustrates this with a small but telling anecdote. Before giving important presentations, he often updates his material right up to the last minute. “I made my presentation when I was in a taxi coming from the airport because I always want to share the latest information,” he said. It’s a detail that reflects not only the pace at NVIDIA, but also the speed at which the entire AI ecosystem is evolving.
For data centers, customers and partners, that pace makes a few things non-negotiable: flexibility by design, continuous learning, and collaboration. Prodanović’s comments point to a market that is moving fast and preparing for a new operational reality.
Subscribe to our newsletter and receive monthly news about LCL, our innovations, our services, our events
