Vertiv Report Explores How AI Is Changing DCs

The rapid acceleration of artificial intelligence is reshaping the global data centre landscape, pushing infrastructure design, power strategies and operational models to their limits. According to the Vertiv Frontiers report, data centre innovation is now being driven by a convergence of macro forces linked to AI workloads, unprecedented scale requirements and the growing complexity of modern compute environments.

Drawing on insights from across Vertiv’s global organisation, the report paints a picture of an industry undergoing structural change as it adapts to the demands of AI factories. These facilities require far higher power densities, faster deployment timelines and more tightly integrated systems than traditional data centres. As a result, the report finds that data centres are increasingly being designed and operated as unified “units of compute,” where power, cooling, IT and software infrastructure must work together as a single system rather than as separate components.

One of the most significant forces highlighted in the report is extreme densification, largely driven by AI and high-performance computing workloads. Traditional power distribution architectures, which rely heavily on hybrid AC/DC systems and multiple conversion stages, are struggling to keep pace as rack densities rise. The findings indicate a clear shift toward higher voltage DC power architectures, which reduce electrical losses, simplify power conversion and better support the extreme power requirements of AI-driven environments. As standards mature and equipment becomes more widely available, higher voltage DC is expected to play a central role in enabling scalable, energy-efficient AI data centres.

The Vertiv Frontiers report also points to gigawatt-scale growth becoming the new norm rather than the exception. Data centres supporting AI are being deployed faster and at much larger scales than ever before, placing pressure on utilities and power grids. In response, operators are increasingly exploring on-site power generation and microgrid strategies to overcome power availability constraints. Extended energy autonomy, once focused primarily on backup resilience, is emerging as a strategic necessity, particularly in regions where grid expansion cannot keep pace with AI demand.

Another key finding from the report is the evolving nature of AI deployment itself. While hyperscale data centres have dominated early investments to support large language models, Vertiv notes a growing shift toward distributed AI. For many organisations, especially those in highly regulated sectors such as finance, healthcare and defence, data residency, security and latency concerns are driving the need for private or hybrid AI environments. This trend is reinforcing demand for flexible, high-density power and liquid cooling systems that can be deployed both in new builds and through the retrofitting of existing facilities.

Speed has also emerged as a defining requirement in the AI era. The report highlights the increasing use of digital twin technology to virtually design, simulate and optimise data centres before physical deployment. By integrating IT and critical digital infrastructure into virtual models, operators can accelerate deployment timelines and reduce time-to-token for AI workloads by as much as 50%. Digital twin-driven design is positioned as a critical enabler of the gigawatt-scale buildouts needed to support future AI innovation.

Cooling is another area undergoing rapid transformation. As AI workloads push thermal limits, liquid cooling is becoming mission-critical rather than optional. The report finds that AI itself is now being used to enhance liquid cooling systems, enabling smarter monitoring, predictive maintenance and adaptive control. These AI-enhanced cooling solutions have the potential to improve reliability, maximise uptime and protect high-value AI hardware in increasingly demanding operating environments.

Overall, the Vertiv Frontiers report underscores that AI is not simply adding incremental demand to data centres but fundamentally redefining how they are designed, powered and operated. From power architectures and energy strategies to digital twins and intelligent cooling, the findings suggest that future-ready data centres will need to be more integrated, autonomous and adaptable than ever before as AI continues to scale globally.