Cooling the Future: The Evolution of Data Centres to Support AI Workloads

Written by David Watkins, solutions director at VIRTUS Published 2024-05-13 09:20:34

In the ever-evolving landscape of technology, artificial intelligence (AI) has emerged as a transformative force, reshaping the way businesses operate and innovate. As AI workloads become increasingly prevalent, not only among industry giants but also across diverse enterprise businesses, the onus is on data centre providers to adapt to these new and demanding requirements. Much like the seismic shift witnessed with the widespread adoption of cloud technology in the early 2010s, we are now at the cusp of another revolutionary wave, where data centres will play a pivotal role in supporting the surge in AI workloads and the wider digital economy.

At VIRTUS, we have long been working on mitigating this challenge, working closely with customers in a collaborative effort. We understand that this transition requires a comprehensive approach, acknowledging the unique challenges posed by the influx of AI workloads. As we move towards a future where AI becomes integral to most industries, it is crucial for data centre providers to align their infrastructure with the dynamic needs of this burgeoning technology.

A cornerstone of this evolution is the imperative to address environmental concerns. In an era where sustainability is at the forefront of every sector, access to truly renewable power at a vast scale is paramount for AI ready data centres. Our new Wustermark megacampus in Berlin is strategically positioned adjacent to one of Europe’s largest onshore wind farms, exemplifying our commitment to minimising its environmental impact. By harnessing the power of renewable energy, we not only support the global shift towards sustainability but also ensure the reliability and scalability required for AI workloads.

The evolving landscape of data centre operations in response to the growing demands of AI is already evident. Whilst data centres have proven their capability in supporting High Performance Compute (HPC) deployments, the unique characteristics of AI workloads present a new set of challenges. Unlike traditional workloads, the demands of AI can be hundreds of kilowatts per rack in their density, with rapidly fluctuating demand. This necessitates innovative solutions including the need to adapt the design of data centres to support the additional weight and cooling characteristics of these substantial AI deployments. These types of unique requirements that AI presents to data centres further underscores the need for a holistic approach to infrastructure design.

Liquid cooling, in various forms, is emerging as a key component in the adaptation of data centres to support AI workloads.  Technological advancements that were once considered 'cutting edge' are rapidly becoming necessities in cooling AI workloads. Immersive cooling and direct-to-chip cooling are two such innovations that are gaining prominence. These technologies, which just a few years ago seemed to be only required for a few select deployments, are now integral to meeting the demands of AI workloads efficiently.

Moreover, the design of data centres must facilitate the integration of standard networking equipment in close proximity to AI deployments. This calls for the availability of multiple cooling technologies to maximise efficiency and create a cohesive environment where the synergy between AI workloads and supporting infrastructure is optimised. You can find a great example of how VIRTUS facilitated this solution for a customer .

The data centre revolution fuelled by AI is not a distant prospect - it is happening now. As the demand for AI workloads continues to grow, data centre providers must be at the forefront of innovation, adapting their infrastructure to meet the unique challenges posed by this transformative technology.  The challenge lies not only in deploying these technologies but also tailoring solutions to meet the unique requirements of individual customers. Collaboration and innovation will be key drivers of success and data centre providers must engage with customers on a case-by-case basis, understanding their unique AI requirements and tailoring solutions accordingly. The ability to provide flexible, scalable and efficient infrastructure will be the hallmark of data centres that lead in the AI era.

As we embrace evolving requirements, finding the right combination of cooling technologies becomes imperative. At VIRTUS, we are committed to empowering the future, ensuring that our data centres not only support but drive the evolution of AI technology, setting new standards for efficiency, sustainability, and performance.