Skip to content
NLEN
Illustration: Grid congestion and AI data centers: state of the queue

Grid Congestion and AI Data Centers: The State of the Connection Queue in the Netherlands

By Ivo Donker — compiled with AI assistance (Claude & Gemini)

In Data centers and AI in the Netherlands the emphasis was on siting and computing power; here it is about the electricity connection as the bottleneck. The rise of large-scale artificial intelligence requires enormous amounts of electrical energy. This places unprecedented pressure on the Dutch power grids, which are struggling with structural grid congestion.

To understand what this means for national infrastructure, it is necessary to take a close look at the current queues maintained by grid operators. This article maps out the state of affairs based on official reports checked on 2026-08-07.

Check date and validity: These figures were checked on 2026-08-07. Because the electricity grid waiting lists are constantly changing due to new publications from grid operators, readers should consult the original sources for the most up-to-date daily status.

What has changed since last time

The dynamics surrounding the power grid and the accommodation of heavy energy users have not stood still in recent months. Anyone following developments will see that the policy and operational frameworks have been tightened:

The Size of the Waiting Lists at Regional Grid Operators

The pressure on the medium-voltage and low-voltage grid in the Netherlands can be read from the semi-annual reports of the joint grid operators. According to the publication Stand van de Uitvoering (semi-annual publication 2025, published by Netbeheer Nederland in September 2025 and announced via the news release of 06-10-2025), regional grid operators had a total of 14,044 unique withdrawal requests on the waiting list, representing a power demand of 9,131 MW. In addition, there were 8,539 feed-in requests with a combined capacity of 4,591 MW.

These figures illustrate that it is not only the withdrawal of power for intensive processes that is problematic, but that feeding sustainable energy back from solar parks and wind turbines is also running up against limits. Anyone wanting to dive deeper into the relationship between power consumption and the computing units themselves can turn to the background piece on AI and energy: why computing power and power consumption matter, which breaks down the theoretical foundations of this energy consumption.

TenneT and National Transport Capacity

On the high-voltage grid of national grid operator TenneT, the situation is of a different order of magnitude. The same semi-annual publication from Netbeheer Nederland (September 2025, released 06-10-2025) showed that the national grid had 212 unique withdrawal requests, with a staggering total power demand of 38 GW. In addition, there were 161 feed-in requests on the list with a capacity of 34 GW.

In the later Landelijk Actieprogramma Netcongestie, whose Stand van de Uitvoering dates from March 2026 and was submitted to the Tweede Kamer in April 2026 (Kamerstuk 2026D15801), it emerged that at the end of 2025, 210 unique withdrawal requests were still on TenneT's waiting list. This mainly concerned industrial large consumers, large-scale battery projects, and data centers. For comparison: the average connection time for small consumers in 2025 was around 45 weeks for a new connection and 21 weeks for a modification. To understand how this gigantic power demand relates to the physical components, the analysis on energy consumption of AI data centers: reading the figures is a valuable addition.

The Hardware Race and the Power Demand of AI Chips

Behind the gigantic power demand of modern data centers lies an intensive technological race. AI chips determine how much computing power a data center requires — see the race for AI chips and computing power for a detailed analysis of this development. As the complexity of foundational language models increases, the density of the required semiconductors per square meter of rack space rises exponentially. This leads to local thermal and electrical peak loads that push traditional cooling systems and transformer stations to their limits.

The ACM Prioritization Framework and the Position of Data Centers

A crucial development in the grid congestion landscape is the entry into force of the ACM prioritization framework. This framework was established by the Autoriteit Consument & Markt on 01-01-2026 and has been fully in effect since 01-07-2026. It sets out a strict order in which grid operators may allocate scarce grid capacity:

As of 01-07-2026, small consumers without priority will also be placed on the waiting list in designated congestion areas, which previously only applied to large consumers. A transition period applies until 01-01-2027.

In this strict system, data centers are explicitly not on the priority list, as also emerged from the interpretation in NRC on 17-02-2026. TenneT spokespeople have confirmed that in practice, freed-up or scarce grid capacity is first reserved for social priorities such as housing construction and healthcare institutions, effectively freezing the queue for commercial data centers.

Regional Bottlenecks and the Situation at Enexis and Stedin

The regional differences in grid congestion are substantial. Enexis Netbeheer reported in its 2025 Annual Report (published in spring 2026) the ambition to reduce the large business waiting list, as of the reference date 01-01-2026, by at least 25 percent over the course of 2026. However, customers who end up on these lists face realities in which they sometimes have to wait 5 to 10 years for a definitive connection. To ease the pain, flexible grid use at Enexis delivered 542 MW of additional capacity during that period.

On the Stedin side, an update on grid capacity from July 2026 showed that as of 01-07-2026, new congestion areas for withdrawal have been established, including at the Europoort 25/10 kV and Wellebrug 25/10 kV stations in the Rotterdam port area and Voorne aan Zee. This leads to so-called "stacked congestion", because these areas also face restrictions on TenneT's underlying and overlying grid.

Flexible Grid Use and Congestion Relief

Grid operators face the immense task of making smarter use of the existing electricity grid. Because building new high-voltage stations and laying heavy cables takes years of permitting and physical work, there is strong focus on flexibilization. Within the current ACM framework, market parties willing to temporarily dim their energy consumption or store it locally are given the opportunity to be connected faster or retain capacity.

The concept of congestion relief revolves around flattening consumption peaks. By using large-scale battery systems, local energy storage, and flex contracts, companies can operate at times when the grid is less heavily loaded. The 2025 Annual Report of Enexis (published in spring 2026) showed that such flexible grid applications and congestion relievers already accounted for as much as 542 MW of additional, directly usable capacity in practice. This shows that smart contracting and power management form an indispensable buffer while awaiting structural grid reinforcement.

What the Queue Means for Data Center Projects

For data center developers in practice, the overload of the power grid has far-reaching consequences. Inventories from the Landelijk Actieprogramma Netcongestie (announced via the Stand van de Uitvoering of March 2026 and recorded in Kamerstuk 2026D15801) show that data centers form a substantial and visible group on the waiting lists of TenneT and regional grid operators. Because data centers are commercial large consumers, they fall outside the categories that receive priority under the ACM prioritization framework.

In practice, this means that when scarcity arises at an overloaded substation or when capacity temporarily becomes available, grid operators give priority allocation to social and primary functions. Concretely, think of the expansion of housing construction sites, hospitals, and vital public services. Commercial data center projects dependent on new, heavy connections are therefore kept at the back of the line, unless they can demonstrate that they fully spare the grid through advanced flex contracts or their own generation and storage capacity.

How to Repeat the Measurement Yourself

Checking the accuracy and currency of grid congestion figures requires a structured approach, since grid operators and regulators periodically publish new datasets. Anyone wanting to verify the national and regional figures, taking into account the check date of 2026-08-07, can use the following public and official sources:

Investments in Grid Expansion and the Future of the Infrastructure

To accommodate the enormous long-term demand for power, the joint grid operators are pumping billions into expanding and reinforcing the electricity grid. According to Netbeheer Nederland's press release of 06-10-2025 (based on the Stand van de Uitvoering of 2025), grid operators already invested more than €8 billion in grid expansion in 2024. These annual investments are expected to rise to more than €15 billion by 2040.

These capital injections are necessary, but have a physical lead time that takes many years. Digging cables, building new transformer stations, and obtaining permits lag behind the exponential growth of the digital economy.

Local Considerations and Measuring Energy Consumption

The pressing situation on the energy grid forces both tech companies and organizations working locally with models to be creative. Anyone facing strict restrictions on the public power grid is increasingly looking at decentralized solutions or more efficient hardware. For organizations that want to know exactly what their own equipment consumes, practical support is available; read how you can measure the energy consumption of your local LLM setup to get a grip on the actual kilowatt-hours in practice. This helps optimize the setup of data centers and office installations within the applicable restrictions.

For organizations considering setting up their own computing power in-house or at a business park, it is also necessary to take these infrastructural conditions into account. A detailed overview of the conditions for this can be found in the technical guide on IT infrastructure for local LLMs, which discusses the practical requirements for power supply and cooling.

The tension between the rapid rise of artificial intelligence and the physical limits of the Dutch power grid will be decisive for the siting climate of data centers in the coming years. The market and the government are continuously balancing between economic innovation and safeguarding public basic services.