What Are Data Centres?
Data centres have become one of the most important pieces of infrastructure of the digital economy.
They power cloud computing, financial transactions, streaming services, social media, scientific research, government systems and, increasingly, artificial intelligence.
Yet despite their importance, remarkably few people know what actually happens inside a data centre — or what physical resources are required to keep our apparently virtual world running.
The word cloud does not help.
The cloud is not somewhere in the air. It consists of physical buildings filled with servers, processors, storage systems, networking equipment and cooling infrastructure.
And all of this requires electricity, produces heat and, depending on the cooling technology and electricity supply, may require substantial amounts of water.
What does a data centre actually do?
At its simplest, a data centre performs three fundamental functions:
it stores data, processes data and transfers data.
When we send an email, make an online payment, stream a film, upload photographs, access business software or use an AI application, computers somewhere have to perform the work.
Those computers are servers.
Large data centres can contain tens of thousands of them, operating continuously and connected through extremely fast networks.
The information stored and processed can include corporate records, financial transactions, scientific datasets, government information, websites, videos, photographs, medical information, communications and personal data.
So yes, our personal information forms part of this enormous digital ecosystem.
But it would be incorrect to describe data centres simply as facilities designed to store personal information or control citizens. They perform an enormous range of legitimate and increasingly essential functions.
There is, however, a legitimate question about the concentration of this infrastructure:
Who controls the infrastructure on which an increasing part of society — and its data — depends?
Then came artificial intelligence
Artificial intelligence is changing the scale of the discussion.
Modern AI systems require enormous computational capacity.
Training large models involves vast numbers of mathematical operations performed by specialised processors, while running those models for millions of users requires continuing computational capacity.
This creates a physical chain that is easy to overlook:
AI → computation → electricity → heat → cooling.
And, in some cooling systems:
AI → computation → electricity → heat → cooling → water consumption.
AI may appear entirely digital on our screens, but its environmental footprint is unmistakably physical.
Where does all that electricity go?
This is one of the most interesting aspects of the discussion.
A server uses electricity to perform calculations, move information and operate electronic components. Ultimately, almost all of that electrical energy becomes heat.
That heat cannot simply remain inside the building. Servers operate within defined temperature ranges and must therefore be cooled continuously.
The heat has to go somewhere.
Cooling systems transfer it away from the processors and ultimately release it into the surrounding environment, unless some of the waste heat is captured and reused — for example, in district heating systems.
This does not mean that data centres are the principal cause of global warming, nor that the heat released by them explains extreme temperatures such as a 45°C summer day.
That would overstate the evidence.
Their contribution to global warming primarily depends on the greenhouse-gas emissions associated with the electricity they consume. Their direct waste heat can be relevant locally, particularly where large concentrations of computing infrastructure exist, but it is not the main driver of global climate change.
Nevertheless, the sheer quantity of energy involved makes the heat question worthy of far more public attention.
How much electricity are we talking about?
The European Commission's Joint Research Centre estimated electricity consumption by EU data centres at approximately 45–65 TWh in 2022.
One terawatt-hour equals one billion kilowatt-hours.
That means roughly:
45–65 billion kWh of electricity in a single year.
And demand is expected to increase as cloud computing expands and AI becomes embedded in more services.
This is therefore no longer merely an IT question.
It is an energy-infrastructure question.
Why do data centres need water?
Not every data centre uses water in the same way.
Some facilities rely largely on dry cooling or closed-loop systems. Others use evaporative cooling, in which water absorbs heat and is subsequently evaporated.
That distinction matters.
Water that evaporates has not merely been withdrawn temporarily.
From the perspective of the local water system, it has been consumed and is no longer immediately available for another local use.
There is also an indirect water footprint.
Electricity generation itself can require water. Thermal power stations — including nuclear and certain fossil-fuel plants — may use substantial quantities of water for cooling.
Consequently, looking only at the water meter at the entrance of a data centre does not necessarily reveal its complete relationship with water.
How much freshwater do European data centres consume?
This is where the picture becomes surprisingly unclear.
There is no single, comprehensive and easily accessible public European database showing, for every major data centre:
how much freshwater it withdraws, how much it actually consumes, where that water comes from and which cooling technology is being used.
European reporting requirements for larger data centres are becoming more extensive, which should improve the available information.
But public transparency remains incomplete.
That becomes increasingly difficult to justify when freshwater itself is under pressure.
A digital economy in a water-stressed Europe
Across Europe, drought, low river flows and pressure on groundwater resources are becoming increasingly important policy issues.
The Netherlands — a country internationally famous for water management — has a formal system for prioritising water use during serious shortages: the verdringingsreeks, or water allocation hierarchy.
That fact changes the context in which we should discuss data centres.
If governments are preparing to decide which societal functions receive priority when freshwater becomes scarce, it is reasonable to ask how large industrial water users fit into that equation.
And it leads to a question that reaches well beyond data centres:
What do we want to use our available freshwater for?
Drinking water?
Nature?
Agriculture?
Industry?
Energy production?
Digital infrastructure?
Artificial intelligence?
These are not anti-technology questions.
They are questions about the allocation of finite physical resources.
The issue is not whether we need data centres
We do.
Hospitals, banks, governments, universities, businesses, communications networks and countless other essential services depend on digital infrastructure.
A modern society without data centres is scarcely imaginable.
The more relevant question is therefore:
Under what conditions should we build and operate them?
If cooling technologies are available that dramatically reduce freshwater consumption, should new water-intensive facilities still be permitted in regions already experiencing water stress?
If a facility consumes enormous amounts of electricity, should its waste heat be discarded, or should heat recovery become an integral part of planning?
And if governments expect citizens, farmers and businesses to accept restrictions during periods of water scarcity, shouldn't major industrial water consumption be fully transparent?
The cloud has a physical footprint
Perhaps that is the most important point.
For years, digitalisation has been presented in language that makes it sound almost weightless.
Cloud.
Virtual.
Digital.
Artificial intelligence.
But underneath those words lies an immense physical infrastructure.
It occupies land.
It requires metals, concrete and sophisticated electronics.
It consumes electricity.
It generates heat.
It requires cooling.
And some of it consumes freshwater.
As AI rapidly increases the amount of computation performed around the world, we should stop treating digital infrastructure as though it exists outside the physical environment.
It does not.
The digital world is physical.
And before Europe expands that infrastructure further, we need much greater transparency about its real resource footprint — not because technological progress should stop, but because responsible technological progress begins with understanding what it actually costs.
Geen opmerkingen:
Een reactie posten