Intelliscale: Meeting AI’s Rising Demands in Data Centres

CyrusOne has introduced its Intelliscale data centre platform designed to tackle the growing power and infrastructure demands driven by AI and high-density computing.

The Intelliscale platform supports rack densities up to 300 kilowatts (kW), presenting a scalable, space-efficient solution that contrasts with the vast footprints of traditional facilities.

With the increasing enterprise demand for AI models' training and deployment, conventional data centres are becoming obsolete.

The shift in data density poses spatial challenges and leads to thermal loads that outdated cooling systems can no longer manage.

Modernising data infrastructure for AI

CyrusOne acknowledges that legacy data centres cannot meet the requirements of contemporary AI and machine learning tasks.

Jim Roche, Senior Vice President of Engineering at CyrusOne, elaborates on how Intelliscale represents a transformation in data centre design. “The evolution of the data centre has been one that's over the last three years has changed drastically.

“If you look back just three years ago, average rack densities were sub 5KW per rack. What we're seeing now, customers coming to the table — they're not asking for 10KW, double that — they're not asking for triple that. They're asking for densities that look like 50KW, 100KW and 300KW per rack.”

“So the challenge is, it's not just moving air around in a data centre anymore. It's how do we approach this holistically and change the design of a data centre to be able to meet the needs of the customer.”

The Intelliscale design focuses on optimising physical space, with the potential to reduce the data centre footprint by up to 75% depending on the use case.

The platform is capable of supporting deployments over 100 megawatts while ensuring reliability and operational adaptability essential for AI-driven workloads. Intelliscale is inherently future-ready, incorporating advanced cooling solutions.

Cooling innovations for the AI era

A key feature of Intelliscale is its ability to integrate both air and liquid cooling technologies.

This adaptability allows customers to expand without requiring extensive infrastructure modifications.

Jim talks about the design facilitating seamless technology transitions, elaborating on the importance of enabling data centres to switch from air to liquid cooling, allowing for scalability and expediting time-to-market through engineering repeatability.

“When we were first designing Intelliscale, we understood that not everybody’s in the market for liquid, but every one of the customers that we talked to said they see a future of it,” he says.

“The thought was: can we design it, make it financially viable to be air in the beginning and if need be, pivot to liquid later on at some time?”

“So when we think about design and build, repeatability is a big thing. We like that philosophy. The more repetitions you get, if you’re doing it the right way, the better you’re going to get at it.

“Our thought with speed to market on Intelliscale was just that — we’re going to take the same components, the same structure, try to work within the same confines, the same building footprint and have it assembled and put together by people who are used to doing it, who have done this one, two, ten, twenty times for us, where that becomes a process.”

The computational demands of AI tasks like model training and data preprocessing require high memory bandwidth, rapid interconnects and large-scale parallel computing. These demands are amplified with increasing data volumes.

Intelliscale's flexible cooling and configuration capabilities ensure customers can accommodate fast-evolving hardware without costly redesigns or downtime.

Building on engineering expertise

Jim emphasises that CyrusOne leverages extensive engineering expertise to develop next-generation facilities.

“One of CyrusOne’s biggest evolutions is our engineering team.

“Not too long ago, we outsourced engineering. What we started seeing was this change in how designs were going to work. We really started looking at the need for people in-house. We need people who are at the table, who are sitting as CyrusOne employees, designing these data centres.

“Liquid in a data centre is not new. What has changed is liquid at scale. We have to keep in mind that these components need to be something that is proven in the industry, readily available and backed by the service provider we’re getting it from.”

Combining engineering consistency with innovation in cooling and density management, CyrusOne aims to deliver high-performance infrastructure that evolves with evolving workload demands.

Its focus on scalable, energy-efficient and modular design meets emerging data centre trends, emphasising sustainability and uptime while maximising compute output per square metre.

As AI amplifies computing needs, solutions such as Intelliscale are vital to the data centre industry's response. CyrusOne's strategy positions it uniquely to support high-density deployments with operational and engineering reliability globally.