SRF meldet zwei Dinge gleichzeitig:
-
Rechenzentren verbrauchen bereits heute 6–8 % des Schweizer Stroms, bis 2030 gelten 10–15 % als realistisch.
-
Gleichzeitig warnen Stromerzeuger vor einer Versorgungslücke bis 2050 (VSE-Index: 69 von 100 Punkten).
Nun in typischer Bünzli-Manier gefragt:
- Wenn der Strom in Zukunft knapp wird, warum geben wir dann ausgerechnet Rechenzentren und KI den Vorrang?
Keine Krankenhäuser, keine öffentlichen Verkehrsmittel, keine Haushalte – sondern Serverhallen für KI-Brainrot, Werbung und Content-Spam?
- Wer trägt das Risiko?
Die Gewinne sind privat, der Netzausbau, die Winterlücke und die Versorgungssicherheit werden von uns allen bezahlt.
- Warum gibt es für alles Regeln – außer hier?
Minergie-Standards für Gebäude, Vorschriften für Heizungsanlagen, CO₂-Ziele überall. Doch die Rechenzentren dürfen weiter wachsen, während die Produzenten gleichzeitig vor Stromengpässen warnen.
Ernsthafte Frage:
Ist es angesichts der absehbaren Stromknappheit wirklich sinnvoll, einen zweistelligen Prozentsatz unserer Stromversorgung dafür zu reservieren – oder brauchen wir endlich klare Prioritäten?
Salopp, ich würde für Rechenzentren ein eigenes Stromnetz inklusive eigener Kraftwerke (wie bei der SBB) vorschlagen, gestaffelt nach Priorität bei Stromausfall. Krankenhausakten usw. hätten Vorrang vor der KI-Generierung für Facebook usw.
Quellen:
Electricity gap by 2050 – and we keep building AI / data centres?
byu/BezugssystemCH1903 inSwitzerland
Von BezugssystemCH1903
12 Kommentare
Because the amount of power consumed by AI is completely misrepresented in the media due to the hype and we simply need these datacenters for our interconnected world, regardless of AI.
Let’s close some more powerplants, what can possibly go wrong?
More cheaper electricity is good. Electricty should not be a scarce resource. We should start building new dams and power plants now.
In the last 20 years the power consumption in switzerland has stayed essentially flat. In the same time, the population grew by 20%, GDP per person also grew about 20% and we have started electrifying many things that previously were powered by other energy sources (mainly fossil fuels). So even though we got wealthier and technology progressed (with all the data centers that come with it), higher efficiency more than made up for this.
I would say the goal is to prioritize what drives the economic wealth of the country. If data centers do more than AI brainrot, then it’s worth it, because economic health is what makes the health system efficient at the end.
[https://fortune.com/2025/10/07/data-centers-gdp-growth-zero-first-half-2025-jason-furman-harvard-economist/](https://fortune.com/2025/10/07/data-centers-gdp-growth-zero-first-half-2025-jason-furman-harvard-economist/)
For house insulation, you can do the same work with less energy. For datacenters you can have datacenters or not have them. That’s the difference.
Also datacenters are very energy efficient all things considered. Paying extra electricity will eat the margins very quickly, so there is a strong incentive to be efficient and the scale to make optimizations worthwhile.
Lastly, if you don’t build them in Switzerland, you will need to depend on datacenters in other countries. The computing needs are not going to vanish.
This is an opportunity to build more power generation.
The Green ppl. Say there is no worry we can cover them with wind and sunny so pls silent Volk the party spoke.
First, we have a claim done by a lobbying body of the power industry. May contain truths but rather bias as well.
Second, yes, electrical power consumption will rise. Not mainly because of AI, but in general with CO2-less heating, E-mobility, population growth and so on.
What to do? I’m a fatalist here, I’ll enjoy popcorn and watch the show. The beauty is that the combination of physics, supply and demand in this case beats politics and ideology by far.
People can talk about the electrical power future all day long, but since no action is taken we have indeed a high chance for power outages in the future.
Because there is such a thing as green and efficient energy!
Is called Nuclear.
Adding generation capacity isn’t cheap (new nuclear power plants ~ 20k/kW), and it doesn’t happen overnight thanks to approval delays, NIMBYism etc. New large scale power consumers should have to carry some of this burden in the form of connection fees or capacity fees.
Most data centers (excluding government / banks / medical) can be located in places where power is more plentiful. E.g. Hetzner has data centers in Finland in addition to their German locations.
It will be enough of a challenge to keep up with added demand from heat pumps and e-mobility.
[https://www.strom.ch/de/nachrichten/vse-stromversorgungs-index-2026-schweiz-verfehlt-gesetzliche-versorgungsziele-deutlich](https://www.strom.ch/de/nachrichten/vse-stromversorgungs-index-2026-schweiz-verfehlt-gesetzliche-versorgungsziele-deutlich)
Beznau will be turned off in 2032 or 2033. Gösgen and Leibstadt could follow in the 2040s unless we extend their lifetime. Building new nuclear power plants is about a 20 year time frame, if the will is there (unlikely) and the numbers add up (unlikely, recent projects have been ruinously expensive).
Renewables can be built more quickly, but it is very difficult to cover large amounts of base load unless you add expensive storage. Rooftop solar doesn’t work when the panels are covered by snow.
Gas peakers with use of the waste heat for „Fernwärme“ sound like the most likely way out. Heat pump power demand correlates well with heat demand of district heating.
* If electricity is going to become scarce in the future, why are we prioritising data centres and AI of all things?
Let’s make 1 thing clear, electricity ISN’T SCARCE WE MAKE IT SCARCE
The serious answer is no, it is not sensible at all.
The only reason this is happening right now is because AI is in a bubble and everyone and their mother is putting billions into its ecosystem. With cheap energy still vastly available this can keep going for a while, but I’m more than happy to bet that predictions of datacenters taking up 10-15% of electricity by 2030 are wrong, even more so as we need to electrify everything that currently relies on fossil fuels (house heating, transportation, heavy industry like cement, steel, etc)
Once electricity production actually becomes a constraint for real you’ll start seeing people taking decisions on where its usage is most important, and it certainly won’t be chatgpt.