New modular data centre design aims to keep pace with AI’s growing power demands

Share this article

A new prefabricated data centre platform has been launched to address one of the most pressing challenges in the age of artificial intelligence – how to deploy high-density computing capacity quickly, efficiently, and at scale.

Vertiv’s OneCore system, unveiled yesterday, combines power, thermal, and IT infrastructure into a single factory-assembled unit designed to reduce the complexity and time involved in building large-scale facilities. The company says the approach is intended to serve high-performance computing (HPC), AI, and hyperscale data centre markets where the demand for capacity is accelerating.

The system is available for enterprise, colocation, sovereign, and neocloud environments, with the aim of helping operators move from design to operational readiness more quickly by streamlining logistics and minimising on-site labour.

Meeting the infrastructure challenge for AI

The growth of AI has placed unprecedented demands on data centre infrastructure, particularly around power density, cooling, and speed of deployment. AI workloads typically require much higher rack densities and more energy-intensive configurations than conventional IT applications, making traditional construction methods less adaptable.

Vertiv says OneCore addresses these challenges by using prefabricated building blocks, such as its SmartRun whitespace fit-outs, housed within a Vertiv-provided steel shell. This allows for parallel manufacturing of electrical and mechanical systems, compressing build times and supporting consistent cost and quality outcomes.

Designed for flexibility, the system can be configured for between 96 and 944 racks and supports scalable power capacity from 5MW to 50MW in a single block. This is intended to accommodate both mixed IT loads and extreme rack densities, with cooling and airflow designs optimised for environmental control in demanding AI and HPC environments.

Viktor Petik, senior vice president of infrastructure solutions at Vertiv, said the goal was to standardise key components while preserving flexibility to adapt to future needs. “The challenge isn’t just designing for today’s needs but building an adaptable foundation for the future,” he said.

Designing for efficiency and resilience

The OneCore platform incorporates Vertiv’s existing cooling and power technologies, including liquid cooling, advanced heat rejection systems, uninterruptible power supply units, and scalable power management. It also offers multiple redundancy options for both fluid networks and electrical distribution, as well as broad voltage compatibility to meet regional standards.

The units are designed to operate in a wide range of climates, from -20°C to 55°C, and to maintain availability during maintenance or upgrades through concurrent maintainability features. Integrated management via the Vertiv Unify system allows operators to monitor and control power, cooling, and IT systems centrally.

Vertiv says the system can support customer sustainability strategies by improving efficiency and reducing power requirements, while its modularity allows facilities to adapt to evolving technology standards without major redesigns.

The launch reflects a wider trend towards modular, prefabricated infrastructure in response to the speed at which AI and high-density computing needs are developing. With AI models becoming larger and more computationally intensive, the ability to deploy capacity rapidly is becoming a competitive necessity for both cloud and enterprise operators.

By consolidating power, cooling, and IT systems into a single, scalable module, Vertiv is aiming to help operators keep pace with that demand while managing costs, build schedules, and operational complexity. For data centre operators facing the dual challenge of AI’s growing energy appetite and the need for rapid scalability, approaches like OneCore signal a shift towards more flexible and industrialised methods of infrastructure delivery.

Related Posts
Others have also viewed
Into The madverse podcast

Episode 32: Your GPUs are waiting for the network

AI performance is not decided by GPUs alone because the network now determines how effectively ...

Why the biggest challenge is not cooling AI but predicting tomorrow’s heat

The AI era has pushed data centre cooling into the spotlight, but the real challenge ...

AI factories will succeed or fail on infrastructure not algorithms

The rapid expansion of artificial intelligence has shifted attention away from software alone towards the ...

The AI infrastructure race is measuring the wrong thing

The AI industry has become fixated on counting GPUs, announcing ever larger clusters and measuring ...