Since the millennium, data centres have been popping up like mushrooms all over the globe. As cloud technology proliferates, providers of data storage and compute power become vital asset holders in the digital economy.

But while essential to our day-to-day lives, data centres are curiously opaque. This is partly an issue of accessibility – unlike power stations, you're rarely, if ever, able to visit one on a guided tour. And they have the kind of top-tier security that you might expect from a Jason Bourne movie.

The other issue is that hard facts about data centres can be hard to come by. Here's a challenge: try to find out how many data centres Microsoft runs. Without going past the first page of Google, it could be 16, 47 or "over 200".

What's more, many data centres are "shrouded in secrecy, or international marketing manipulation".

This quotation is taken from a fascinating long-read article by Sebastian Moss which attempts to establish where the world's largest data centre is. He tries to find out with a combination of official data, satellite images and good old-fashioned investigative journalism ("we reached out to data center technicians working at the facility, but have yet to hear back").

The main takeaway from the article is that those listicles about the "world's biggest data centres" should be taken with a pinch of salt. Poorly sourced – or not sourced at all – they're the blog equivalent of churnalism.

So what do we know? Well, Moss gives the following answer:

• Meta's Prineville data centre campus is the largest data centre cluster.
• The largest single data centre building is probably the 11-storey Singapore data centre, also owned by Meta.

Some existing data centre sites are going to be expanded – and when they are, could top the charts. These are:

• Switch's Citadel Campus in Nevada
• The Range International Information Hub in Langfang, China
• Huawei's Campus in Guizhou province, China

Our difficulty in knowing which will be the biggest lies not only in the fact that they've not been finished – it also lies in the fact that these projects' marketing campaigns could be plain wrong.

What is a "hyperscale" data centre?

Data centres come in all sizes. To take one almost at random, the Edge 8 data centre in Birmingham, England is a 63,055 sq ft facility that can house up to 2,000 racks.

Tech giants like Amazon, Meta, Apple, Google, IBM and Microsoft naturally take up more space. They build massive facilities at hyperscale – or rent them from colocation and data centre development companies like Equinix and Digital Realty.

Hyperscale data centres differ from their smaller counterparts by being highly scalable. As demand grows, so too do the centres – and to achieve this, servers, racks, cooling systems and the rest are mass-produced for quick and cheap installation.

Some facilities are single-tenant, meaning that one company (IBM, say) runs a cloud application across a single site. Others are multi-tenant, meaning that the infrastructure is shared between clients – an increasingly common option.

Case study: Microsoft's Columbia data centre in Quincy, Washington

However many data centres Microsoft operates, one of the biggest is the Columbia Data Centre in Quincy, Washington, which has been running since 2007.

Taking a brief virtual tour of the campus, the size and scale of the operation has the same dizzying effect as Jorge Luis Borges' short story "The Library of Babel" – only instead of infinite hexagonal rooms full of books, you have endless stacks of servers, row on row, building on building.

Some stats to get you in the mood: the campus contains over half a million servers, storing more data than 50,000 Libraries of Congress. And it consists of more than 20 buildings – each one big enough to house two commercial aeroplanes.

Data centres are, of course, physical buildings vulnerable to everything from power outages to heatwaves. To this end, Columbia has extensive backup in place in the event of a power outage – more than 19,000 battery cells and 140 generators. Each generator could power the equivalent of 3,000 homes.

Microsoft's empire of data centres is continuing to grow, with Madrid next on the list.

What next for data centres?

The most controversial aspect of these mammoth data centres is their environmental impact. In December 2021, the Financial Times reported that "already, data storage and transmission […] are slurping up about 2.5% of global electricity usage" – and that by 2030, 15% of Denmark's energy use will be needed to power its data centres.

One leftfield response to this has been Microsoft's Project Natick – the world's first underwater data centre. So far, a prototype has been deployed and successfully retrieved – albeit now covered in algae, barnacles and other sea stuff – from the seafloor off the Orkney Islands.

Compared to on-land data centres, subsea centres are far more sustainable – partly because they don't need the same cooling infrastructure.

You can see this electricity-guzzling cooling in action in this video from Google. Different data centres use different methods – some use air conditioning units, others cool water cooling.

Microsoft also claims that its data centre's backup generators are on their way out – to be replaced by the end of the 2020s either with hydrogen-powered generators or "new advanced fuel cells".

Another issue facing data centres is the increasing frequency of heat waves. In July 2022, Google and Oracle data centres in London had to be shut down because of the record-breaking heat. The cooling systems were overheated and overwhelmed, leading to website outages across Europe.

Conclusion

Data centres are forever being built and expanded. Without them, the internet as we know it wouldn't be able to function – but it's not always easy to find concrete information about them.

Whatever happens, you can be sure that they're going to keep getting bigger.

Ascend Cloud Solutions is a cloud consultancy and migration firm headquartered in Cork, Ireland. If you're looking for expert guidance, please don't hesitate to get in touch.