Have We Already Lost the AI Economy?

The Netherlands wants to remain a leading knowledge economy. We want to attract AI companies, stimulate innovation and increase our productivity. We want to become less dependent on American and Asian technology, and we want Dutch and European technology companies to grow here.

Against all these ambitions stands a fundamental question that is rarely asked out loud: do we actually have the infrastructure to make them possible? Not the strategy, not the talent, not the capital, but the physical foundation on which all of this must run.

This contribution is not aimed only at the real estate and IT sectors. It is intended for government, ministries and policymakers, for grid operators, municipalities and provinces, for energy companies, technology companies, employers, investors, developers and entrepreneurs. Because this issue concerns them all.

AI Doesn't Run on Ambition. AI Runs on Infrastructure.

AI is often discussed as if it were purely software. A model, an algorithm, a smart application. But every AI query, every model, every cloud application and every data stream ultimately requires physical infrastructure.

AI requires data, computing power, electricity, connectivity, data centers and digital infrastructure. This makes AI not just a technological issue. It is at the same time an energy issue, an infrastructure issue, a real estate issue and a spatial planning issue.

That is the central contradiction. Our economic ambitions are growing exponentially, while a large part of our physical infrastructure cannot keep pace at the same rate. That tension is not a theoretical risk. It is already visible right now in the Dutch electricity grid.

The Netherlands Already Has Grid Congestion

In large parts of the Dutch grid, available transport capacity is fully allocated. Grid operators publish capacity maps on which entire regions are marked red, both for feed-in and for consumption. Companies, projects and even residential areas are on waiting lists for a heavier connection, and those waiting times stretch to years.

The cause is not a shortage of electricity as such, but a shortage of capacity to transport that electricity. The grid was designed decades ago for a different reality. Now electrification, solar parks, heat pumps and growing industry all demand space on that same grid simultaneously. The investments grid operators need to make in the coming years to reinforce it run into the tens of billions, and execution is constrained by staff, materials and permits.

Grid congestion is therefore not only a problem for housing or industry. It could ultimately also become a limitation for data centers, offices, laboratories, campuses, logistics, charging infrastructure, high-tech companies, AI companies and the digitalization of existing businesses. The question that arises: what happens if our digital economy grows faster than our energy infrastructure?

How Much Energy Will AI Actually Require?

The electricity consumption of data centers, cloud infrastructure and AI is growing rapidly internationally. International energy organizations assume that global electricity consumption by data centers will increase substantially in the coming years, driven by the shift toward compute-intensive AI applications. A classic search query and a conversation with a large language model differ considerably in energy demand per operation.

It is wise to make a distinction here. There is current electricity demand, which is already substantial in itself. There is expected growth, which remains uncertain because efficiency and demand are both increasing. And there is the difference between data centers in general and specifically AI-related computing power, which is rising the fastest in relative terms.

Sensational predictions do not help. What does help is a sober observation: the trend points upward, and that growth comes on top of a grid that is already becoming congested. This leads to a simple but confronting question: where should all that extra electricity come from?

Can We Solve This With Wind Turbines?

Suppose AI, data centers, electrification of mobility, heat pumps and industry all require more electricity. How much additional production capacity would we actually need? A thought experiment helps clarify the order of magnitude.

A modern offshore wind turbine generates a substantial amount of energy annually, and a complete wind farm represents many terawatt-hours per year. If the additional future demand from Dutch data centers were to amount to several terawatt-hours, that would on paper require a considerable number of large wind turbines or multiple wind farms to cover that annual production.

A wind farm delivers energy over the course of a year. A data center demands power at every single moment of that year. That difference is exactly where the calculation breaks down.

And there lies the core of the misunderstanding. Annual energy production is not the same as capacity that is continuously available on demand. Wind is variable. Data centers run 24 hours a day, seven days a week. That is why not only production, but also grid capacity, storage, flexibility, interconnection and dispatchable power are all relevant. So the question is twofold: how many wind turbines must we build to keep up with our digital ambitions, and is energy production even the only problem if the grid still has to transport that energy afterward?

What About Nuclear Energy?

Nuclear energy is often mentioned as the answer to the need for dispatchable, continuous power. That is not unreasonable: a nuclear plant delivers stable power regardless of wind and sun. But the claim that a nuclear plant simply takes thirty years deserves some nuance.

In European practice, it is wise to distinguish between political decision-making, permitting, financing and actual construction time. The pure construction period of recent projects is substantial, but it is mainly the years preceding it, decision-making, procedures and financing, that stretch out the total timeline. Realistically, we are talking about many years to over a decade before new large-scale capacity actually delivers electricity.

This raises a much more interesting question. If new large-scale energy infrastructure needs ten, fifteen or twenty years, but AI can fundamentally develop within two or three years, aren't our economic and infrastructural timelines diverging completely?

That is one of the key themes here. AI develops in months. Real estate in years. Energy grids in decades. The question is not which of these three can move fastest, but how we bring three such unequal speeds together without the slowest one blocking the rest.

From Grid Congestion to Data Congestion

Within IT-Label's vision, one concept is important, explicitly intended as a way of thinking rather than as an established technical fact: data congestion. It is not a formal, nationally recognized term like grid congestion. It is a way of drawing a parallel that is too rarely made.

We now know exactly what happens when more electricity is demanded than the local infrastructure can transport. But are we already asking the same question about our digital infrastructure? What happens when companies start processing exponentially more data, while buildings, connections, internal cabling, WiFi networks and technical rooms fail to grow along at the same pace?

The idea, then, is this: after grid congestion, digital capacity could well become the next infrastructural bottleneck. We explored this idea earlier in our article on whether real estate can keep up with the growth of AI, and the question is becoming more urgent by the day.

The Forgotten Infrastructure: Commercial Real Estate

Here lies the translation that is often missing. AI companies do not ultimately work only in data centers. AI is used from offices, business premises, universities, hospitals, laboratories, government buildings, campuses and millions of workplaces. It is precisely there that a blind spot arises.

We talk about energy labels, ESG, CO2, grid congestion, data centers, AI strategies, digital autonomy and cybersecurity. But do we ask often enough: can our commercial real estate actually handle the AI economy? The smart building technology that modern organizations rely on assumes a building that is physically ready for it.

A 2005 Office With a 2030 Organization

Take a recognizable example. An organization rents a beautiful office. Good energy label, good accessibility, sufficient parking, attractive appearance, a fair rental price. On paper, everything checks out.

Then two hundred employees move in who work daily with cloud applications, video conferencing, large datasets, AI assistants and an increasing number of real-time applications. Suddenly it becomes relevant which fiber connection is in place, whether there is redundancy, how old the internal cabling is, what the WiFi infrastructure can handle, how cybersecurity is arranged, how technical rooms are set up, whether expansion is easily possible and whether the power supply supports future IT equipment.

The conclusion is unpleasant but clear. A building can structurally still last for decades while at the same time already being digitally outdated. For this phenomenon we use the term digital ageing of real estate. Anyone who moves into a space without checking these points runs into surprises, as we describe in our article on what the IT-Label tells you in advance about a shell space.

Our Knowledge Economy Has a Physical Foundation

The Netherlands rightly calls itself a knowledge economy. But knowledge today is processed, stored, shared and developed almost entirely through digital systems. That means a knowledge economy today is also a data economy.

A data economy can only function with sufficient energy, sufficient grid capacity, data centers, fiber, cybersecurity, cloud infrastructure, good digital connections, digitally capable buildings and digitally capable workplaces. Our articles on network redundancy show how vulnerable an organization becomes as soon as one link in that chain is missing.

Perhaps digital infrastructure is to tomorrow's economy what roads, ports and railways were to yesterday's. Invisible until it is missing, and then suddenly decisive for everything that runs on top of it.

A Critical Question for The Hague

The Netherlands is investing billions in energy grids. We are developing AI strategies. We talk about digital autonomy. We want to retain innovative companies and we invest in cybersecurity. But one question remains unanswered: who monitors the digital readiness of our built environment?

Who knows how much commercial real estate is actually suitable for increasingly digital organizations? Who knows which buildings have sufficient connectivity? Who knows where digital infrastructure has become outdated? Who brings together the combination of energy, data and real estate? And above all: which ministry feels responsible for this?

Is digital real estate an IT issue, an economic issue, a real estate issue, an energy problem or a spatial planning issue? Probably all of these at once. And that is exactly the risk. When an integrated issue has no clear owner, it falls between departments. We explored this earlier in our article on IT labels and government policy.

Do We Need a Digital Infrastructure Policy for Buildings?

The message is not that the Netherlands has definitively lost this battle. The message is sharper: if we do nothing now, infrastructural limitations could hold back our AI ambitions faster than any shortage of talent or innovation ever could.

That is why it is worth examining whether the Netherlands, alongside energy policy and AI policy, should also think more explicitly about the digital quality of the built environment. Not as additional regulatory burden, but as a way of making visible something that currently simply is not measured.

The Possible Role of IT-Label

It would be implausible to present IT-Label as the solution to the Dutch energy or AI challenge. It is not. IT-Label positions itself as a possible missing measurement instrument within one essential part of the puzzle: commercial real estate. What we do not measure, we cannot easily improve.

We want to make visible what digitally often remains invisible. Think of connectivity, fiber, redundancy, internal data cabling, WiFi, technical rooms, cybersecurity-related building facilities, continuity, digital capacity, smart building infrastructure, expandability and the division of responsibilities between tenant and landlord. In our overview of the classification from PREMIUM to SHELL, we translate this into understandable levels.

This creates a picture of how digitally future-proof our commercial buildings actually are. Not as a judgment of good or bad, but as objective information on which owners, tenants and policymakers can base their decisions. What a classification means is explained on our page about what the IT-Label is.

A Bigger Ambition

Imagine that within a few years the Netherlands knows not only how energy efficient its buildings are, but also how digitally future-proof they are. Imagine that municipalities, investors, tenants, developers and government structurally factor digital infrastructure into development, renovation and housing policy.

Then the Netherlands could distinguish itself. Not only as a country with good digital connections, but as the first country to systematically prepare its built environment for an AI-driven economy. That is not a technical detail, but a strategic choice that fits the ambition to remain a leading knowledge economy.

The Uncomfortable Choice

The world is not waiting for the Netherlands to expand its electricity grid. AI does not wait for our permitting procedures. Technology companies do not wait twenty years for infrastructure. Capital and talent can relocate, and they do.

That is why the real competitive battle may not be only about who develops the best AI, but about which country has the energy, the digital infrastructure and the buildings to actually house that AI economy.

Have we already lost the AI economy? Probably not. But we could lose it if our ambitions grow digitally faster than our infrastructure can physically follow. The Netherlands is a knowledge economy. A knowledge economy runs on data. Data runs on digital infrastructure. Digital infrastructure requires energy. And ultimately everything comes together somewhere: in a building.

Alongside the question of how much energy we can supply and how much AI we can develop, we must therefore dare to ask a third question: can our built environment actually handle tomorrow's digital economy? Anyone who wants to make this question concrete for a specific building or portfolio can take a good first step by visiting our page on how the IT-Label works to see how this digital quality is made visible. Because the future of real estate must be not only sustainable, but also digital.