Cloudy with a chance of colocation…

The age-old argument within the IT department started with a simple choice: build vs buy. Should companies invest in building their own data centre, or rent space from a colocation provider? 

Next came the cloud vs colocation debate, when organisations had to weigh up the cloud benefits of scalability, agility and lower upfront costs against colocation’s control, predictable operating expenses and enhanced security.

But what does the landscape look like now? It’s certainly more complex, with an increasing array of options for IT departments to consider. 

The Evolving IT Landscape: 2024 and Beyond

Rewind a decade to 2014, when the digital world seemed a simpler place. The financial services sector dominated data centre deployments with many banks owning entire data centres or leasing large (at least considered large at the time) amounts of colocation space. Such deployments ran at steady state, using modest amounts of power, particularly when network racks and a tape library or two are factored in. 

Fast-forward to today, and whilst the financial sector’s IT needs haven’t diminished, hyperscale cloud providers now house much more of their IT infrastructure as hybrid cloud solutions have become the standard. Where banks once might have required 5MW of IT Load, cloud providers now demand upwards of 50MW just to get a data centre investment started.

Changes are afoot

The world has changed.  Generative AI is here and financial institutions are now developing their own AI systems, prompting a new question: should these systems be hosted on public cloud platforms or are they better suited to private clouds within colocation facilities where control and security can be more tightly managed? Should they even revert to the self managed on-premise solution as compute power scales to unprecedented levels?

This shift is, of course, not exclusive to the financial services sector. The AI phenomenon and the rise of GPU-as-a-service has opened up a whole new world for enterprise and is transforming business strategies everywhere. Whether organisations are using AI technology in CRM systems, delivery and travel or customising and generating marketing messaging, retail or research, the compute power required is driving capacity requirements to significantly increase. 

Universities and research organisations, which require highly specialised infrastructure to support their supercomputers, are increasingly opting for colocation services. Whilst public cloud providers have the resources to support the training for AI workloads, the significant increase in capex infrastructure for AI capable servers, coupled with the rising power costs, means that businesses are seriously reviewing their strategies to enable scale at the lowest total cost of ownership. Although the cloud provider option means a speedier route to these services and lower upfront capital commitments, as these services scale, so potentially can the costs dramatically increase. 

No one-size fits all

As transformative as AI is - and it is - the underlying principles and drivers for business remain the same. The difference is the magnitude of the costs and the speed of change. As the public cloud providers race to be the first to truly monetise AI and incorporate it into their ubiquitous service platforms, businesses are still facing the same conundrum: Capex vs Opex, Cloud vs colocation, or the best of both in the form of hybrid cloud.

Whatever path organisations take, colocation remains at the heart of the IT strategy in one form or other. The challenge now is to secure capacity whilst it’s available. As good as AI is, it hasn’t yet solved the power challenges it contributes to. 

The availability of scalable, hybrid data centre space has never been more important, and to that VIRTUS Data Centres is opening LONDON14 in Spring 2025, specifically designed to support this type of requirement.