Vertiv forecasts five trends reshaping AI data centres

Vertiv expects shifts in power architectures, wider adoption of digital twins and increased use of liquid cooling to define the next phase of data centre design and operations as artificial intelligence drives rapid growth in computing demand.

The company has published its latest Vertiv Frontiers outlook, which sets out how operators are responding to extreme rack densities, gigawatt-scale campuses and an industry view of the data centre as a single unit of compute.

The report outlines four macro forces. These are extreme densification, gigawatt scaling at speed, data centre as a unit of compute, and silicon diversification. Vertiv links these forces to five technical and operational trends that it expects to shape new builds and retrofits.

Scott Armul, Chief Product and Technology Officer at Vertiv, said these forces are reshaping infrastructure choices across the sector.

"The data center industry is continuing to rapidly evolve how it designs, builds, operates and services data centers, in response to the density and speed of deployment demands of AI factories," said Vertiv chief product and technology officer, Scott Armul. "We see cross-technology forces, including extreme densification, driving transformative trends such as higher voltage DC power architectures and advanced liquid cooling that are important to deliver the gigawatt scaling that is critical for AI innovation. On-site energy generation and digital twin technology are also expected to help to advance the scale and speed of AI adoption."

Vertiv positions these developments as globally relevant but highlights specific implications for operators in Asia, where demand for AI workloads and strict regulatory frameworks in sectors such as finance and healthcare influence infrastructure planning.

Power for AI

The first trend centres on power systems for AI-heavy environments. Most current facilities use hybrid AC/DC power distribution from the grid to IT racks. This structure usually involves three or four conversion stages. Vertiv notes that such designs are under strain as power densities rise.

The report sets out a shift towards higher-voltage DC architectures. That approach reduces current, shrinks conductor sizes and cuts the number of conversion stages. Power conversion moves to the room level instead of at each rack or device.

Hybrid AC and DC systems remain common. Vertiv expects higher-voltage DC to become more prevalent as rack densities climb and as standards and equipment mature. The company also links the trend with on-site generation and microgrids, which it says will encourage further adoption of higher-voltage DC distribution.

Distributed AI

A second trend focuses on where AI inference runs and how it reaches users. Large language models have driven heavy investment in centralised AI data centres. Vertiv expects AI to become increasingly important for businesses in multiple industries, but says delivery models will vary by sector and risk profile.

Highly regulated industries such as finance, defence and healthcare may keep private or hybrid AI environments. They may operate these via on-premise data centres because of data residency rules, security concerns or strict latency requirements. Flexible power systems and liquid cooling in high-density configurations feature in Vertiv's view of how operators might expand capacity for such customers, through new facilities or upgrades of existing sites.

Energy autonomy

The third trend deals with energy strategies. Data centres have long installed short-term on-site generation for backup. Vertiv says ongoing grid constraints and wider power availability challenges are now pushing operators, especially those running AI data centres, toward extended energy autonomy.

The outlook cites increasing investment in on-site generation using natural gas turbines and other technologies. Vertiv describes several intrinsic benefits, but places primary emphasis on the need to address availability risks. The report also references strategies described as "Bring Your Own Power (and Cooling)", which it expects to feature in long-term energy autonomy planning.

Digital twins

The fourth trend is the use of digital twins during design and operations. Vertiv links this to the need to deploy dense AI workloads and GPU-based infrastructure at speed.

Using AI-based tools, operators can create virtual models of data centres as digital twins. Design teams can specify layouts, map power and thermal flows and integrate IT with critical infrastructure in a virtual environment. Vertiv says operators can then deploy these configurations as prefabricated modular units of compute.

The company states that this approach can reduce the time it takes to bring AI capacity online. It says digital twin-led design can cut time-to-token by up to 50 per cent and can support gigawatt-scale build-outs that it expects to accompany further AI growth.

Liquid cooling

The fifth trend addresses thermal management. Vertiv reports that AI workloads and high-density infrastructure are driving faster adoption of liquid cooling across the sector.

The analysis describes liquid cooling as mission critical for a growing number of operators. Vertiv expects AI techniques, combined with additional monitoring and control systems, to influence how such systems run. These tools can forecast potential failures and manage fluid flows and components more actively.

Vertiv says this adaptive approach should increase reliability and uptime for high-value hardware and associated data and workloads. The company also highlights that operators are designing liquid solutions as critical load, with redundancy and controls that support uptime in mixed environments, where air and liquid cooling work together depending on workload and site conditions.

The company operates in more than 130 countries and supplies power management, thermal management and IT infrastructure across data centres, communications networks and industrial facilities. Vertiv expects the outlined trends to shape decisions on future AI-ready infrastructure and on the design of large-scale data centre campuses.