The title may evoke homemade flying bicycles, wobbly boats and other improvised vehicles bravely taking on the water, with more enthusiasm than engineering. Half a century on, a more serious variant is playing out, and the stakes are high. The question is no longer how to get a barrel across a ditch, but where to place the engines of our digital economy: on land, at sea or in the air.
Data centres are the quiet engines of our digital economy. Every payment, every video call, every AI prompt travels past a row of servers in a cooled hall somewhere. For the Netherlands, that is no abstraction. With the AMS-IX, Amsterdam has for years been one of the most important internet exchange points in the world, and that position has helped turn our country into a digital hub for Europe. But it is precisely that lead that is now under pressure.
Why it is straining on land
The brake is not on demand, but on the preconditions. In many places the electricity grid is running up against its limits, and grid congestion means new connections can take years to arrive. On top of that comes the water debate: cooling requires water, and in dry summers that has become a sensitive subject. Add scarce space and public resistance to large hyperscale sites, and you understand why new projects are getting ever harder to get off the ground. The AI wave does not make things any easier, because AI clusters consume many times more power per rack than a classic data centre. So it is only logical that we are looking upward. Literally.
Space as a compute centre
Elon Musk has painted the picture that within a few years it could be cheaper to place AI data centres in orbit than on the ground. The reasoning is seductive in its simplicity. In space the sun shines almost without interruption, so you need no batteries to bridge the night. And because it is bitterly cold and a vacuum, you can radiate heat away without the water cooling and chillers that cost so much energy and water on Earth. Large satellites full of solar panels and AI chips, and the water and grid problem evaporates in one go.
There is another advantage that weighs ever more heavily on Earth: the matter of permits. A data centre on land can easily disappear for years into zoning plans, environmental and water permits, and objection procedures from local residents. In orbit, that whole circus simply does not exist. There are, admittedly, launch licences, international agreements on frequencies and rules against space debris, but there is no local resistance and no years-long permit battle. That saves not only time, but also uncertainty in planning.
Google, too, showed with its exploratory project around chips in orbit that this is not idle bar talk, and start-ups are trying to raise serious capital for it. Even so, there are considerable snags. Launching is still expensive, even though the price per kilo keeps falling. Maintenance in space is virtually impossible: you cannot simply swap out a failed drive. Radiation wears down electronics faster than on Earth. And radiating heat away in a vacuum sounds simple, but is technically the harder part, because you cannot shed it through air or water. Add latency to that, and you understand why this looks promising for now mainly for heavy, batch-based AI training, and not for services that have to respond in milliseconds.
Or into the sea after all?
The alternative lies closer to home: the sea. With Project Natick, Microsoft already showed that a sealed data centre can run perfectly well underwater, with the ocean as free, stable cooling. China has since taken that step to commercial scale with subsea data centres off its coast. The idea is elegant, because cooling is your biggest headache on land, and seawater largely solves that. What is more, much of the world's population lives close to the coast, so the distance to users stays manageable.
But here too there are doubts. Maintaining a capsule on the seabed is laborious, the ecological impact of discharging heat into vulnerable coastal waters is contested, and the power still has to come from somewhere, ideally from offshore wind. For a country with a long coastline and a strong wind sector, however, it is a line of thinking that fits the Netherlands remarkably well.
For now, the ground keeps winning
Are space and sea realistic alternatives? In the longer term, certainly, and it is wise to follow them seriously rather than laugh them off. But for the coming years the most obvious answer remains a sober one: data centres on land, preferably close to the network exchange points where the data already converges. That is where the connectivity is, that is where the scale is, and that is where the real gains are to be made, not with spectacular launches but with more efficient cooling, smarter grid management, residual heat that warms neighbourhoods, and AI clusters that flexibly match their power demand to the supply of wind and sun.
The future of data centres probably lies not in one radical breakthrough, but in a combination. Space for specialist AI training, the sea as a regional cooling solution, and the ground as the backbone that holds it all together. The real question for the Netherlands is not whether we want data centres, but how we build them so that they fit within the limits of our grid, our water and our space.
