Artificial intelligence (AI) is profoundly changing the architecture of data centres, and one of the most visible changes is related to cabling infrastructure. Thus, AI-based data centres need up to ten times as many connections as traditional ones, said Romeo Lazăr, a regional representative for Corning, a US company with optical cabling factories in France, Germany and Poland.

“The wiring difference between a conventional data centre and an AI-optimised one is huge, because of the high density and specific needs for GPU processing,” Lazăr said at the DigitALL 2025 conference organised by Energynomics.

According to him, there are three major types of data centres: enterprise (managed internally by companies), multi-tenant (hosting the infrastructure of several clients, including in Romania) and hyperscale (operated by giants such as Amazon, Microsoft or Google). All of them can be adapted to AI requirements, but the transition involves significant logistical and technical challenges.

One of the critical issues is rack cooling and fast delivery of connectivity equipment. Corning is banking on structured cabling, with durable centre trunks from which patch cord connections are made, replacing point-to-point solutions that, while simple, no longer meet today’s needs.

DOWNLOAD THE PRESENTATION OF ROMEO LAZĂR

“Increasing density, the need for scalability and rapid deployment are the dominant trends in this area. Having factories in Poland allows us to quickly deliver critical components to Eastern Europe,” added the Corning representative.

The “DigitALL 2025” conference was organised by Energynomics, in partnership with reputed organisations such as the CIO Council and EPG, with the support of our partners: ABB, Corning, Datacor, Eaton Electric, Enevo Group, Enevo Group, Procesio, Renomia, script.ai, Sunwire, Vertiv.