Even Bits Need a Place to Work: The Rise of the Data Centre
Even Bits Need a Place to Work: the Rise of the Data Centre
As we’ve discussed before, the office has changed a great deal in the last quarter-century. Our views about that transformation are structured around a broader narrative about the evolution of information and communications (ICT) technologies. We’ll start by laying out that narrative, since it is crucial to understanding the rise of the data centre.
Where did all the paper go?
Thirty years ago or so, electrons started to push atoms out of the workplace, as paper documents and filing cabinets were gradually replaced by digital forms of communication and information storage. Office workers got desktop computers, and the information now displayable on their monitors was mostly stored in private servers, which sat in dedicated rooms on-site. By the early aughts, most workplaces operated with that type of technology set-up. As computer usage proliferated and information storage further digitised, physical ICT infrastructure swallowed up more and more office space. We wound up with crowded, open-plan offices with lots of servers. The paper was effectively transferred out of office filing cabinets and put in office server cabinets.
In the past decade, and particularly within the last few years, that paradigm evolved. While the first phase of the computing revolution saw workplaces swap paper files for digital ones, the second phase delivered significant changes to computing infrastructure itself. Although the volume of stored information has exploded – a major trend in its own right – physical ICT infrastructure consumes much less office space today than it did ten years ago. That’s mostly due to the fact that enterprise computing operations increasingly shifted to the cloud.[1]
Where did all the servers go?
With the introduction of the cloud, computing infrastructure could be hosted remotely, managed by specialised operators, and consumed “as a service” (and therefore upsized or downsized easily) by end users located anywhere. About 70% cheaper than traditional storage,[2] cloud computing presented a compelling value proposition, particularly during a period in which high-growth technology companies were multiplying at a rapid pace. This was because high-growth firms typically (1) aimed to minimise fixed investment in costly hardware; (2) wanted the flexibility to adjust computing capabilities in lockstep with business growth; and (3) tried to keep their teams lean, devoting as few resources as possible to non-core business functions, like technology infrastructure.[3] The sheer volume of content that internet companies increasingly needed to store and deliver also made it infeasible to host that content in-house. As a result, enterprise spending on cloud infrastructure services increased from $1.5 billion in 2010 to $129.5 billion in 2020.[4] As organisations migrated their digital operations to the cloud, they started to abandon the private servers that they had previously crammed into on-premises server rooms. This meant that offices could now fit more people.
But all that physical computing infrastructure did not vanish. It just moved. After all, the cloud doesn’t actually float in the sky – it takes up real space on the ground. More specifically, the cloud is housed within millions of servers that sit inside thousands of specialised warehouses called data centres. We like the title that one Wall Street analyst uses for his semi-annual research report on the sector: “The Cloud Has Four Walls.” Very roughly: add a roof, a raised floor, adequate, redundant power supply, plenty of cooling capacity, network connection points, and thousands of servers, and those four walls form a data centre.
Migration to the cloud led to the physical consolidation of computing infrastructure. When companies relocated their digital operations to the cloud, they usually traded server rooms in their own office buildings for rack space in third-party warehouses,[5] owned primarily by cloud service providers (like Amazon Web Services and Microsoft Azure) or multi-tenant data centre operators (like Equinix and Digital Realty). In short, servers left offices and moved into data centres.
Like many structural trends in real estate, the rise of the data centre is partly a simple compositional story: it’s about the movement of digital information out of private servers in office buildings and into vast “server farms” in data centres. This adjustment was borne out in real estate markets. The data centre sector realised double-digit compound annual total returns between 2014 and 2020, outpacing office returns by a wide margin.[6] These gains were achieved in tandem with a massive increase in supply. Annual capex investment in hyperscale data centres tripled between 2014 and 2020 (from $30 billion to $94 billion).[7] In Europe’s “Tier 1” markets (Frankfurt, London, Amsterdam, and Paris), third-party data centre capacity almost quadrupled between 2010 and 2020.[8]
So, in certain ways, what’s bad for the office is good for the data centre. Some of the forces that have dragged down aggregate demand for office space – in particular, the growing adoption of cloud computing – have pushed up aggregate demand for data centre space.
[1] “Cloud computing” refers to the delivery of infrastructure, platforms, and software via the internet.
[2] Based on the total cost of ownership for a typical enterprise user, as estimated by Credit Suisse in 2018.
[3] It’s not a coincidence that analogous considerations have propelled the rise of flexible offices, as we have noted before: both cloud computing and flexible offices operate by turning the provision of capital goods, which were historically owned or leased by users, into pay-as-you-go services. As a result, they attract customers for similar reasons.
[5] Hardware stored in data centres is held in metal cabinets or “racks”, and customers often effectively rent units of rack space.
[6] Calculated using NAREIT’s total returns indices for office REITs and data centre REITs.