Electricity is a unique energy carrier: energy is transferred by electrons. Its discovery was an evolutionary process, but with Michael Faraday's discovery of electromagnetic induction, the fundamental laws of nature became visible and applicable. An energy carrier is not a fuel; it merely transports energy. That energy can be generated using a variety of fuels: gas, nuclear energy, solar energy, or hydropower. Ultimately, all these sources have one historical origin: the sun. Without the Big Bang and our sun, there would be no Earth, no wind or rain, no fossil fuels, and not even uranium.
Electronics is a specialized field within electrical engineering. Here, we let electrons perform their "tricks" in a carefully designed playground of silicon. This enables mathematical operations: the basis of computers (the engines), software (the rules), and data (the building blocks and end products).
Where we used to talk aboutcomputing centers, with the emphasis oncomputing power, we now talk aboutdata centers. It is not the computing itself, but the data that represents the greatest economic value today. Power plants charge electrons with energy; data centers are the places where those energy-rich electrons are used to process information.
The more information we create and process, the more energy is needed to move electrons and enable them to do their work. As information processing grows in a society, the demand for electrical energy inevitably grows as well. In the era of telex, radio, and television, that demand was negligible. In today's information society, it has become structural and dominant.
Writing this blog, having it edited by ChatGPT, and publishing it on the internet consumes electricity. That energy is used once. During that process, it loses its high-quality form and ends up as low-quality energy: heat.
Energy is never lost, but must always be 'processed' for specific applications. Thousands of degrees or more for steel production, hundreds of degrees for steam turbines in power stations, tens of degrees for home heating. Ultimately, all electrical energy ends up as heat. This is not a policy choice, but a consequence of the laws of nature. This also applies to data centers.
The high-quality, stable electricity that enters a data center leaves the building after use as low-grade heat of several tens of degrees. Sometimes it can still be used for district heating, but otherwise it is useless. The surplus disappears into the atmosphere and space, within the limits determined by the sun, oceans, and atmosphere.
When we demand intensive information processing 24/7—for example, for large-scale AI models—a continuous, intensive energy supply is also necessary. That energy demand is growing exponentially, not linearly. Traditional electricity generation is already under pressure. Due to their intermittent nature, solar and wind power are fundamentally unsuitable for continuous high-power supply.
After all, electricity is difficult to store, especially in the quantities and capacities required for data centers and AI clusters. Anyone who ignores this is ignoring not only technology, but also physics.
This explains why large information processors are increasingly turning to nuclear energy. Nuclear fuel has an extremely high energy density and can supply large amounts of stable electricity day and night. Without electricity, there are no energy-rich electrons. Without energy-rich electrons, there are no powerful chips and processors. And without those processors, there is no intensive information processing such as AI.
This is not ideology, but chain logic.
Right now, there is a growing awareness—albeit reluctant—that our popular energy transition is not designed for the information transition it is supposed to support. On the contrary. In addition to an existing electricity grid, which is already heavily burdened by digitization, we have set up a second infrastructure for intermittent "green" generation.
Two networks. One system. And an industry that demands high-quality energy 24/7.
Furthermore, the costs of this second network are not allocated exclusively to green generation, but are socialized across all users. The result: reliable, continuous electricity is made artificially expensive, while green energy itself remains dependent on that same network.
Anyone who truly wants green energy should also be prepared to bear the full system costs. Not rely on infrastructure designed for stable, high-quality use. What we are doing now is allowing slow traffic on the highway to save on bike paths — which is exactly what large-scale feed-in from solar panels actually means.
Low-density energy is mixed with high-density energy. And everyone pays the price.
Data center operators have understood this for a long time. Their business model stands or falls with reliable electricity. That is why we are seeing a clear trend: data centers are organizing their own energy supply.
Not out of luxury, but out of necessity.
Energy generation and data centers are converging—functionally, technically, and commercially.For example, it was recently announced that Pure Data Centers in Amsterdam is starting a project in Westpoort where the energy supply has already been secured prior to construction via a private substation. It is striking that Amsterdam's largest data center is being fully utilized by a single hyperscaler.
Concentration squared.
Companies such as Microsoft, Google, and Amazon have been moving toward owning their own power plants for some time now. Not because they want to, but because they have no other choice.
Information and energy are inextricably linked. Anyone who takes digitization, AI, and "the cloud" seriously must also dare to talk about nuclear energy, grid capacity, and the laws of nature.
Without high-energy electrons, there can be no high-energy information processing.
And without that recognition, the policy remains stuck in wishful thinking.
