Solar and wind power are likely to grow worldwide at a pace compatible with limiting global warming to 2 °C – but not 1.5 °C, according to an analysis with a new, AI-powered model that the researchers involved liken to a “computational time machine.”
Until now, it’s been hard to say how fast solar and wind power will expand in the future, as rapidly falling costs on the one hand bump up against policy changes, public opposition to particular projects, and delays in getting new installations hooked up to the grid on the other.
The upshot: we’ve been able to calculate what needs to happen with renewable energy to reach climate targets, but not how likely we are to get there.
In the new study, researchers simulated solar and wind rollout in 13,000 virtual worlds, each composed of 150 “countries” similar in size to real-world nations. This enabled them to leverage historical data, especially from early-adopter countries, to sketch out every possible trajectory of renewables development.
Then they trained a machine learning algorithm on those virtual worlds to predict how renewables will develop in a given country, and across the globe as a whole, based on early data. “Our projections show the most likely development of solar and wind power given current trends,” says study team member Avi Jakhmola, a graduate student at Chalmers University of Technology in Sweden.
Overall, the team’s median estimates show that onshore wind is likely to represent about 25% of global electricity supply in 2050, and solar about 20%. These figures are consistent with limiting warming to 2 °C but not 1.5 °C.
At the COP28 summit in 2023, countries pledged to triple renewables by 2030. “The COP28 pledge is within the realm of possibility but would require sustained acceleration across all major regions, comparable to what the European Union is attempting with its RePowerEU plan,” Jakhmola says.
Similarly, expanding solar and wind fast enough to limit warming to 1.5 °C would be challenging but not unprecedented, the model suggests – if we get started now. If we wait until 2030 the required deployment curve gets steeper and steeper.
“It’s also worth noting that our projections are not static, but update as new national data comes in, allowing us to track how the outlook evolves over time,” Jakhmola says.
The researchers created a freely available, interactive online tool to help people visualize the results.
To further validate the new model, the researchers fed in data just through 2015, and asked it to predict what has happened in the 10 years since. The model was remarkably accurate, even besting the International Energy Agency’s own projections. That’s why they dubbed it a “computational time machine.”
The strong performance “gave us real confidence that the approach is capturing something fundamental about the growth of these technologies,” Jakhmola says.
In the future, the model might also be able to help predict the development of other green technologies. “Wind and solar were a natural starting point because they have been used long enough in different countries to give us enough empirical data to work with, but we think the same approach could be applied to technologies like electric vehicles, green hydrogen or battery storage as they mature,” Jakhmola says.
Source: Jakhmola A. et al. “Probabilistic projections of global wind and solar power growth based on historical national experience.” Nature Energy 2026.
Image: © Anthropocene Magazine. AI-generated
