Executive Summary
- Data Centre Insight spoke with Darren Watkins, chief revenue officer at VIRTUS Data Centres, about what it takes to operate data centres with the capacity to support AI at scale.
- Facilities that have been designed for density, scalability, resilience and sustainability will allow for AI to be fun more efficiently while meeting ESG and carbon-reduction targets.
- Modular construction, and flexible cooling and power systems ensure UK data centres remain adaptable and competitive as the use of AI continues to grow.
Across every sector, organisations are racing to embed artificial intelligence (AI) into their operations. But as models scale and training demands intensify, one truth is becoming clear – AI success depends as much on physical infrastructure as digital innovation.
Data Centre Insight spoke with Darren Watkins, chief revenue officer at VIRTUS Data Centres to explore what it really takes to build and operate data centres capable of supporting AI at scale.
Q: Why are so many enterprises struggling to operationalise their AI strategies?
Many organisations underestimate the physical footprint that AI requires. Training and inference workloads consume far more power and generate far more heat than traditional IT. While companies have invested heavily in data science and cloud adoption, their underlying infrastructure often cannot sustain those demands. Research from S&P Global shows that the proportion of businesses shelving AI initiatives has more than doubled in the past year. The issue is rarely software. It’s that legacy environments simply weren’t designed to deliver 50 – 80 kilowatts or more per rack, continuous runtime, or precision cooling at scale.
Q: What makes AI workloads fundamentally different from traditional enterprise IT?
AI changes both the intensity and the profile of demand. Traditional IT systems – finance, ERP, office applications – run predictable workloads with manageable peaks. AI is the opposite. Training large models requires thousands of graphics processing units (GPUs) running for weeks, often without pause, and even in production, inference workloads are highly dynamic.
Those power and thermal profiles can’t be met by standard on-premises data rooms or older colocation halls. Modern AI workloads call for facilities built from the ground up for high density, liquid cooling, and advanced power management.
Q: What are the defining features of an AI-ready data centre?
At VIRTUS, we define AI readiness through five lenses: density, cooling, resilience, scalability, and sustainability.
· Density: Facilities must comfortably support racks drawing tens of kilowatts each, with room to scale further.
· Cooling: Direct-to-chip and immersion cooling systems are now essential for high-efficiency heat removal. Many facilities are moving to hybrid models that blend air and liquid systems to balance performance and cost.
· Resilience: Every layer of the power distribution system needs redundancy and intelligent monitoring to maintain uptime.
· Scalability: Modular hall design allows capacity to grow as AI workloads expand.
· Sustainability: Energy efficiency and heat reuse are built in from the start, ensuring new AI capacity aligns with ESG targets.
These factors combine to create an environment where AI workloads can operate reliably, safely and efficiently.
Q: How is AI changing the way data centres are designed and built?
The shift begins at the design stage. The next generation of data centres is being engineered for much higher rack densities, with reinforced flooring and infrastructure to manage the weight of liquid cooling systems. Electrical rooms are larger and more complex, with additional switchgear and uninterruptible power supply (UPS) capacity to support the higher draw.
We are also seeing a rise in megacampus development – sites designed to deliver hundreds of megawatts of capacity, with modular phases and direct grid connections. This scale allows operators to integrate renewable energy, deliver higher efficiency, and meet the enormous power requirements of AI clusters.
Q: Why is location now such a critical factor for AI performance?
Proximity to data has become a competitive advantage. AI models rely on vast, distributed datasets, and latency directly affects their accuracy and responsiveness. If compute is too far from the data source or the end user, performance degrades.
That’s why many enterprises are now selecting facilities based on connectivity and latency, not just cost. Sites near data exchanges, financial hubs, and population centres are becoming strategic assets. In many ways, the data centre has evolved from a passive utility to an active part of the AI value chain.
Q: Sustainability remains a challenge. How can AI infrastructure grow responsibly?
AI will inevitably increase power demand, so efficiency must improve at every level. At VIRTUS, we’re investing heavily in renewable energy sourcing, waste-heat reuse, and precision cooling systems that adapt dynamically to load.
We’re also using intelligent monitoring and automation to optimise energy use in real time. These advances allow us to support high-density AI environments while meeting our customers’ sustainability goals and national carbon-reduction targets.
Q: How can operators ensure their facilities stay adaptable as AI continues to evolve?
Flexibility is the foundation. AI workloads and regulatory frameworks are changing quickly, so infrastructure needs to evolve with them. Facilities that can move seamlessly from 20 kW to over 100 kW per rack, or expand cooling systems without downtime, will remain future-ready.
Modular construction, scalable density, and advanced management platforms are key. They allow customers to grow capacity incrementally, shift workloads as needed, and maintain compliance without rebuilding from scratch.
Q: What is the strategic takeaway for organisations investing in AI?
AI transformation is as much about infrastructure as innovation. The enterprises moving fastest are those aligning their digital ambitions with physical readiness. Data centres designed for AI are becoming strategic enablers: offering the power, cooling, and connectivity that make advanced compute possible.
As AI continues to reshape the economy, those foundations will define who can innovate at scale, and who cannot.
