Artificial intelligence is accelerating changes in how data centres are designed and operated, pushing the industry towards higher power densities, new cooling technologies and greater energy autonomy, according to a newly released report.
The Vertiv Frontiers report said forces such as extreme densification, gigawatt-scale deployments and the emergence of the data centre as a single unit of compute are reshaping infrastructure requirements as AI workloads grow.
“The data centre industry is rapidly evolving how it designs, builds and operates facilities in response to the density and speed demands of AI factories,” said Scott Armul, Vertiv’s chief product and technology officer.
Rising AI and high-performance computing workloads are placing strain on existing power architectures, which typically rely on hybrid AC and DC power distribution with multiple conversion stages. The report said higher voltage DC architectures could reduce inefficiencies as rack densities increase, particularly as full DC standards mature.
On-site power generation and microgrids are also expected to drive adoption of higher voltage DC systems, as operators face growing challenges around grid capacity and power availability.
The report also pointed a shift towards distributed AI. While large investments have focused on hyperscale data centres supporting large language models, Vertiv said regulated industries such as finance, defence and healthcare may increasingly rely on private or hybrid AI environments hosted on-premise, driven by data residency, security and latency requirements.
Energy autonomy is emerging as another priority. While backup power has long been used for resilience, the report said AI-driven power demands are accelerating investment in extended on-site generation, including natural gas turbines and strategies such as Bring Your Own Power and Cooling.
To speed deployment, data centre operators are increasingly turning to digital twins, the report said. Virtual modelling allows facilities to be designed, integrated and deployed as prefabricated units of compute, potentially cutting deployment times significantly as operators move towards gigawatt-scale builds.
Liquid cooling is also becoming mission-critical as AI workloads increase thermal demands. Vertiv said artificial intelligence could further enhance liquid cooling systems by predicting failures, optimising performance and improving reliability for high-value hardware.
