Although the term artificial intelligence was coined by John McCarthy as early as 1956, it is the past decade that has made it omnipresent. The decisive factor in its breakthrough into everyday life has been ChatGPT, which brought AI from technical environments into daily life.
For AI systems to function, they require a vast amount of data and exceptional high-performance computational infrastructure. Such systems impose vastly different and higher requirements for the environments where they are deployed. Existing data center infrastructure is inadequate for the installation of high-performance systems required for AI operation.
New challenges
AI systems will soon be integrated into specialized business environments — for electricity consumption forecasting, logistics, traffic prediction, market trend forecasting, and customer behavior modeling. The fundamental advantage of AI lies in its capacity to process vast amounts of data at a speed far surpassing human capabilities.
Historically, average power consumption of system cabinets ranged between 5–10 kW. The Nvidia DGX B200 AI system can reach up to 56 kW per cabinet. Some systems require between 130 and 400 kW per cabinet, with projections suggesting requirements may escalate to 1 MW per cabinet in the near future.
Preparing for hybrid data centers
A new paradigm is emerging — the hybrid data center, designed to accommodate both conventional ITC systems and AI-driven systems with significantly higher power demands simultaneously. This necessitates a fundamental shift in design philosophy and the adoption of direct liquid cooling as an imperative. Hybrid centers will feature pre-installed piping networks for technical cooling, allowing seamless integration of new AI systems.
These challenges were the focal point of the NTRsync 2025 conference (March 26, Maribor, Slovenia), attended by Europe most esteemed IT and AI experts and leading global providers of advanced AI-driven solutions.