Energy demand from data centers in Belgium is set to rise sharply in the coming years. New figures from grid operator Elia have once again brought this issue to the forefront. At the same time, an investment of more than one billion euros was announced for a new data center campus on the former Sappi site in Lanaken.
At Datacenter United, we’re seeing the same trend from a different angle. This week, we signed an agreement to provide several megawatts of capacity for AI workloads at our data center in Machelen.
These three developments all boil down to the same thing: AI is beginning to create a very concrete demand for physical infrastructure in Belgium as well. And that requires a lot of energy. We can’t ignore that. But if the debate stops at the question of how much electricity data centers consume, we’re only asking half the question. The other half is at least as important: what is Belgium at risk of losing if the infrastructure behind our digital economy cannot keep pace here?
AI is digital. The infrastructure behind it is not.
AI models may consist of software and data, but the infrastructure on which they run is very tangible:
- Servers
- Electricity
- Connectivity
- Cooling
- Space
- Redundancy
As AI systems become more powerful, the amount of computing power that must be concentrated within a limited physical space increases. This means higher power densities and significantly more heat that must be efficiently dissipated.
The consequences of this are already becoming apparent. According to new projections reported this week, electricity demand from Belgian data centers could rise sharply by 2034. Not all requested or reserved capacity will necessarily be realized in the end, but the trend is clear. The digital economy will require more physical infrastructure.
That future is no longer a prediction. It is being built right now, today.
At Datacenter United, we don’t have to look as far ahead as 2034 to see that evolution. At DC Machelen, DCU has signed an agreement to make several megawatts of capacity available for AI workloads. The planned environment will use liquid cooling to support the high power densities and heat loads of modern AI systems. At the same time, we are already investing in advanced cooling infrastructure for high-density environments in Machelen. Capacity, cooling, and technical infrastructure are also being expanded or modernized at other locations within our Belgian network.
These are not investments based on a hypothetical future. They respond to a demand that is emerging today.
The announced investment of more than one billion euros in a new data center campus in Lanaken also illustrates how rapidly the market is evolving. Moreover, according to current plans, the project would create 150 to 250 jobs at a site where hundreds of jobs were lost following the closure of the former factory. This makes digital infrastructure not only a technological or energy issue, but also an economic one. At the same time, this is still a development that must go through the various stages of planning, permitting, infrastructure, and construction before operational capacity can become available. That distinction is important.
Because the AI challenge is emerging today, not when new infrastructure is completed in a few years.
Belgium must therefore look to the future while also being able to provide sufficient suitable capacity today. New developments are necessary to meet future demand, while existing and available infrastructure is needed for companies that want to realize their AI ambitions today. That should not make the debate on energy any less important. It should broaden the discussion.
What we don't make possible here doesn't disappear. It just moves elsewhere.
AI development won’t stop just because Belgium lacks sufficient capacity, energy, or infrastructure. Companies will continue to need computing power. Investors will continue to build data centers. AI models will continue to be trained and used. The question is where. That is precisely where a risk lies that is much harder to see than a data center’s electricity consumption. When infrastructure is built elsewhere, investments, technological expertise, digital ecosystems, jobs, and future economic activity may also shift along with it.
What we don't make possible today, we can't just bring back tomorrow.
As a result, digital infrastructure is becoming more than just a technical issue. It is having an increasingly significant impact on our economic competitiveness and digital sovereignty. How dependent do we want to become on infrastructure outside Belgium and Europe for technology that is becoming increasingly fundamental to our businesses and economy? And how much control are we relinquishing as the physical layer underpinning our digital economy moves further and further away from us?
That doesn't mean: data centers at any cost.
The answer can never be that every megawatt must automatically be allocated or that every data center request must be fulfilled. On the contrary. Precisely because energy is valuable and not infinitely available, we must set higher standards for the infrastructure it powers.
How much computing power do we extract from the available capacity? How efficiently is the infrastructure cooled? How is energy consumption monitored and optimized? Can the infrastructure support higher densities without a disproportionate increase in consumption? What role can renewable energy, waste heat, and flexibility play?
At Datacenter United, we strive to make that responsibility measurable. We monitor PUE at the site level to identify efficiency gains in cooling, airflow, and operational settings. In Antwerp and Oostkamp, we have a combined installed solar capacity of 821.2 kWp. Our climate targets have been validated by the Science Based Targets initiative, and through our sustainability reporting, we also make our own energy consumption and environmental impact transparent. Not because this will eliminate energy demand, but because growth is only responsible when efficiency and environmental impact are integral parts of infrastructure decisions.
Every megawatt has to work harder.
Today, the discussion about data centers is too often framed as if energy consumption and digital growth were two conflicting interests. That’s too simplistic. AI will require more energy. There is little doubt about that. The challenge, therefore, is not to pretend that this energy demand does not exist. The challenge is to ensure that every available megawatt is used as efficiently as possible and creates as much digital, economic, and social value as possible.
This requires modern infrastructure, new cooling technologies, transparency regarding impact, and collaboration between data center operators, energy companies, grid operators, governments, and technology firms. Belgium currently boasts strong connectivity, technical expertise, a central location, and a growing digital ecosystem. But that position is not guaranteed.
Investment decisions are being made today. AI infrastructure is being planned today. Capacity is being allocated today. Other European regions aren’t standing still while Belgium debates. That’s why we must not reduce the debate to the question of how much energy a data center consumes. We must also dare to ask what that infrastructure makes possible.
Which companies can grow here? What technology can we develop here? What knowledge and expertise are we building here? What jobs and economic activity are we creating here? And how much control do we want to retain over the AI infrastructure on which our future digital economy runs?
The energy costs of digital infrastructure are immediately apparent. The economic value it creates—across businesses, jobs, innovation, and technology—is much harder to capture in a single figure. We may not realize the cost of not building it until the investments, technology, jobs, and economic activity have already shifted elsewhere.
Words from this article
Data centers and AI involve a lot of technical terms. That’s why we’ll briefly explain a few of these concepts in this article.
Data center: a secure physical location where IT systems, such as servers and network infrastructure, are housed. Among other things, data centers provide the power, cooling, connectivity, and security needed to ensure the reliable availability of digital applications and data.
AI workload: a computing task required to develop or use artificial intelligence. Examples include training an AI model or having an existing model process a task. Such applications can require large amounts of computing power.
Computing power: the capacity of computers and servers to perform calculations and process data. The more complex a digital application—such as AI—is, the more computing power it typically requires.
Megawatt (MW): a unit of power. One megawatt is equal to one million watts. In a data center, MW is used, among other things, to indicate how much electrical power is available to support IT infrastructure.
Power density: the amount of electrical power used within a given physical space. AI systems can concentrate a great deal of computing power into a limited area, thereby increasing power density and the need for efficient cooling.
High-density environment: a data center environment in which a large amount of computing power and electrical power is concentrated in a relatively small area. Such environments place higher demands on power supply and cooling, among other things.
Liquid cooling: a cooling technology that uses liquid to dissipate heat closer to the heat source. This can enable efficient cooling for powerful AI systems that are densely packed together.
Redundancy: the provision of backup or duplicate systems for critical infrastructure, such as power supply, cooling, or connectivity. If one component fails, another system can take over to help ensure continuity.
Connectivity: the network connections that enable a data center to exchange data with other locations, networks, cloud platforms, and the Internet.
PUE (Power Usage Effectiveness): A commonly used metric for measuring a data center’s energy efficiency. PUE compares the data center’s total energy consumption to the energy that goes directly to the IT equipment. The closer the value is to 1, the smaller the proportion of energy that goes to supporting infrastructure such as cooling.
Digital infrastructure: the physical and digital systems that enable digital services, including data centers, servers, networks, fiber-optic connections, cloud infrastructure, and other technology on which digital applications run.
Digital sovereignty: the extent to which a country or region can maintain control over its digital infrastructure, technology, and data, and is not entirely dependent on infrastructure or technology outside its own region.