AI pressure drives changes in data centers
What is behind the accelerated growth of artificial intelligence and the unprecedented pressure on data centers?
A new report reveals the invisible forces that are redefining how the digital infrastructure of the future is designed and sustained.
The race to support the advancement of artificial intelligence is not only fought in algorithms and models but also in the infrastructure that makes them possible. With this premise, a new Vertiv report warns that data centers are undergoing a profound transformation, driven by extreme demands for energy, speed, and density, which are reshaping their design, operation, and scalability.
The Vertiv™ Frontiers report, based on the company’s global experience, identifies the major macro forces and technological trends shaping the present and future of data centers—from preparing for AI workloads to leveraging digital twins and adopting adaptive liquid cooling.
Scott Armul, Vertiv’s Director of Product and Technology, explains:
“The data center industry continues to evolve rapidly in how it designs, builds, operates, and services these facilities, responding to the density and deployment speed demands of AI factories. We are seeing intersecting technological forces, including extreme densification, driving transformative trends such as high-voltage direct current (DC) power architectures and advanced liquid cooling, which are essential for scaling to gigawatt capacities critical for AI innovation. On-site power generation and digital twin technologies will also help accelerate AI adoption.”
The document expands on Vertiv’s previous annual predictions about data center trends. It outlines four macro forces accelerating innovation: extreme densification driven by AI and high-performance computing (HPC) workloads; rapid scaling to gigawatt-level capacity; viewing the data center as an integrated computing unit; and silicon diversification, which requires infrastructures to adapt to a growing variety of chips and architectures.
Based on these forces, the report identifies five key trends that are already directly impacting the data center ecosystem:
1. AI Readiness
The growth of AI workloads is testing traditional power distribution models. Many data centers still rely on hybrid AC/DC schemes, involving multiple conversion stages and efficiency losses.
Faced with increasing per-rack density, the report highlights a gradual shift toward high-voltage DC architectures, which reduce current, conductor size, and conversion stages while centralizing power transformation.
On-site power generation and microgrids are also emerging as factors that will drive this transition.
2. Distributed AI
Although much of the initial investment in AI data centers has focused on supporting large-scale language models, the report emphasizes that AI inference will become increasingly critical for business operations.
In highly regulated sectors such as finance, defense, and healthcare, requirements for security, low latency, and data protection could drive private or hybrid AI environments, supported by local data centers.
In this context, flexible, high-density power systems and liquid cooling emerge as key enablers, both for new facilities and for upgrading existing infrastructures.
3. Accelerating Energy Autonomy
Limited energy availability has become a structural challenge for the industry. While on-site generation has historically served as a backup solution, the report warns of a shift toward extended energy autonomy strategies, especially in AI-focused data centers.
Investment in technologies such as natural gas turbines responds not only to resilience needs but also to the urgency of ensuring continuous power supply. In this scenario, approaches like Bring Your Own Power—and cooling—gain relevance as part of long-term planning.
4. Digital Twin–Driven Design and Operations
The complexity and density of AI factories require deploying infrastructure on increasingly shorter timelines. The use of digital twins, supported by AI-based tools, allows virtual mapping and specification of data centers, integrating IT and critical infrastructure into modular, prefabricated designs.
According to the report, this approach can reduce time per token by up to 50% and will be crucial for efficiently achieving gigawatt-scale expansions required by future AI advancements.
5. Adaptive and Resilient Liquid Cooling
The adoption of liquid cooling has accelerated significantly due to the thermal demands of AI infrastructure. At the same time, AI itself is beginning to play a role in optimizing these systems through advanced monitoring, intelligent control, and failure prediction.
The report notes that this combination could result in more robust cooling systems, with higher reliability and uptime for critical hardware and high-value workloads.
|
|
