Forecast: AI-optimised fusion reactors and direct carbon removal from the atmosphere will solve the CO2 problem. The traditional environmental movement, which relies on austerity, will become politically irrelevant.
The AI revolution consumes energy on a scale that surprises even experts. The International Energy Agency (IEA) estimates global electricity consumption by data centres in 2024 at around 415 terawatt-hours -- approximately 1.5 per cent of worldwide electricity consumption [1]. By 2030, this figure is expected to more than double to over 945 TWh -- equivalent to the entire electricity consumption of Japan.
| Key figure | Value |
|---|---|
| Data centres worldwide 2024 | 415 TWh |
| Forecast 2030 | 945 TWh |
| Growth rate | 15% per year |
| Share USA + China | 80% |
| Swiss total consumption | 58 TWh |
A single training run of a large language model like GPT-4 consumes as much electricity as 10,000 Swiss households in a year [2]. And that is just the training. Ongoing operation -- billions of queries per day -- comes on top.
Nuclear fusion, for decades "thirty years away", is coming within reach through AI-assisted plasma control. The computing power for simulating fusion plasma, which in 2010 would have required decades, has been reduced to months by AI.
Key players:
Forecast: The probability that nuclear fusion will be commercially usable by 2050 exceeds 70 per cent. When fusion arrives, energy scarcity becomes a memory, not a problem.
In parallel with fusion, AI is dramatically improving the efficiency of existing solar and wind installations:
Scenario: Instead of solving the CO2 problem exclusively through consumption reduction, AI-controlled direct carbon removal from the atmosphere (Direct Air Capture) becomes economically viable [5]. AI-controlled agriculture -- swarms of small robots tending fields plant by plant -- eliminates pesticides and herbicides. What is organic luxury today becomes the standard supply.
Switzerland generates around 60 per cent of its electricity from hydropower and decided by popular vote in 2017 against new nuclear power plants. For a country that wants to be an AI hub, this is an uncomfortable truth: whoever lacks energy will become a spectator to a revolution happening elsewhere.
Microsoft has announced plans to bring decommissioned nuclear power plants back online. Google and Amazon are investing in small modular reactors [6]. The message is unmistakable: Whoever wants to control the future of AI must first solve the energy question.
The dominant political narrative of the 2020s -- climate emergency and economic restrictions -- will be overtaken by technological progress. This is politically explosive but historically consistent: technology has always solved scarcity problems more reliably than appeals to austerity.
[1] International Energy Agency (IEA): Electricity Mid-Year Update 2025. Paris, 2025.
[2] S&P Global: Data Center Power Demand. 2025.
[3] ITER Organization: ITER -- the way to new energy. iter.org, 2025.
[4] Helion Energy: Company Profile and Investors. helionenergy.com, 2024.
[5] Climeworks AG: Direct Air Capture Technology. Zurich (Swiss company, world leader).
[6] Microsoft: Announcement on Three Mile Island Reactivation. September 2024.