How Vertiv is Powering AI-Ready Edge Infrastructure

AI is transforming data infrastructure wholesale, and with it, energy systems themselves. According to Andrea Ferro, VP Power and IT Systems EMEA at Vertiv, the pace of AI adoption is forcing data centre operators to rethink the fundamentals of power density and efficiency.

McKinsey projects AI-ready data centre capacity will rise 33% annually between 2023 and 2030, while Goldman Sachs forecasts AI could trigger a 165% leap in total data centre power demand by that year.

"It's the speed of AI adoption scaling and its impact on resource use," Andrea says.

Today’s high-performance racks already consume 120–132kW, yet by 2027 that could rise to 600kW. It is expected that that figure will eventually reach 1MW for future generations.

“This isn’t just about scaling existing infrastructure – it's about rethinking power delivery, thermal management and system integration," Andrea explains.

“We’re no longer dealing with isolated server deployments but with fully integrated AI factories that require increasing amounts of power to enable computing at scale.”

New energy patterns emerge at the edge

This evolution is most evident at the edge, where the shift from centralised AI training to distributed inference is changing how energy is consumed. As workloads move closer to users and their data sources, energy use becomes more distributed. As such, efficiency gains can depend on granular control.

“From a technical perspective, training workloads can tolerate higher latency and benefit from centralised, high-throughput architectures," says Andrea. "Inference, particularly for agentic AI applications, demands microsecond response times with consistent performance."

The result is a growing network of high-density, energy-hungry nodes at the edge which are small in footprint but heavy in power requirements.

Each one needs reliable power distribution, local cooling and intelligent energy management to achieve both performance and efficiency.

Efficiency challenges multiply as power scales

The edge AI market is forecasted to expand from US$20.78bn in 2024 to US$66.47bn by 2030, with energy performance now the critical variable.

“Enterprises need local processing for speed, control and efficiency. Edge data centres help with that,” Andrea explains.

“But there’s another factor: energy efficiency," he adds. "Processing data at the edge reduces the energy costs of data transmission and allows for more granular power management. With data centre energy consumption set to double by decade's end, this efficiency gain is becoming crucial.”

Unlike traditional edge applications, physical AI systems bring new demands: they must operate in industrial or outdoor settings while matching the energy density of a data centre rack.

“This creates unique infrastructure challenges,” Andrea says. “These systems often need to operate in harsh environments such as manufacturing floors, outdoor installations and mobile platforms, while maintaining the computational power of a data centre.”

The rise of AI is not about abandoning the cloud, but rather about rebalancing where power is consumed.

“AI needs a hybrid approach that leverages both centralised and distributed capacity,” says Andrea. “Training large models still happens in cloud clusters, but inference is increasingly happening at the edge.”

This rebalancing allows organisations to optimise performance, manage energy costs, and meet regional energy and data regulations. The key, Andrea emphasises, is aligning each workload with its best-fit energy and compute environment.

Addressing high-density power delivery

As the industry chases greater density, Vertiv’s engineering teams are working on the limits of safe and efficient power delivery. Many modern systems now operate between 120 and 132kW per rack, with next-generation architectures targeting 600kW by 2027.

“This creates a cascade of technical challenges,” Andrea says. “Power delivery requires 33kW DC power shelves and 1400A busbars – infrastructure that was not required or imagined for pre-AI traditional data centres.”

Liquid cooling technologies like Vertiv’s CoolChip CDU 100 handle these escalating heat loads while maintaining energy efficiency. At such scales, cooling itself becomes a major part of the power equation.

Energy resilience and adaptability are key priorities for operators modernising legacy sites. Retrofitting attracts interest for its lower disruption, but older power distribution systems were not designed for today’s AI loads.

“Upgrading from traditional power delivery to 33kW DC power shelves and 1400A busbar systems often requires electrical infrastructure changes,” Andrea says. Meanwhile, integrating liquid cooling into air-based systems demands significant mechanical and control adaptations.

Modular systems like Vertiv’s 360AI reference architectures – from 88kW to 115kW – offer a bridge solution. These integrated designs allow facilities to add high-density energy capacity without starting from scratch and can cut deployment time by up to 50%.

Designing for sustainability and future energy systems

As AI scales, so do its energy demands. However, it can also open doors to opportunities for sustainable innovation. “This goes beyond traditional metrics like Power Usage Effectiveness (PUE) to include circular economy practices like heat reuse, water recycling and end-of-life equipment management,” Andrea says.

Regulations and customer expectations are pushing operators to measure and disclose their energy use more transparently.

For Andrea, the guiding design principle has to be adaptability: building systems that evolve with future energy realities.

“The main principle is designing for adaptability and scalability, not just current requirements,” he says. “AI is evolving rapidly, and infrastructure needs to accommodate today's applications as well as tomorrow's breakthroughs.”

Engaging partners early and using standard reference designs supports both energy efficiency and speed of deployment. “Build with change in mind as it is the only constant,” Andrea concludes.

“The decisions we make today about edge infrastructure will enable or constrain the next decade of AI innovation.”