Wind and solar power may be heading for a faster global expansion than most official forecasts have projected, according to a machine-learning framework that learned its predictions from 13,000 simulated national energy histories. The system, called PROLONG, was described in an April 2026 study published in Nature Energy, and its median outlooks for renewable capacity through mid-century sit above the ranges that governments and climate bodies have long treated as ambitious targets.
The finding lands at a moment when forecasters are already scrambling to keep up with real-world installations. The International Energy Agency has revised its renewable projections upward in each of the past several annual outlooks, and global solar additions alone have roughly doubled since 2021. PROLONG’s authors argue that even those revised numbers may still undercount the momentum already visible in country-level deployment data.
How PROLONG works
Rather than optimizing for cost curves or GDP assumptions the way traditional integrated assessment models do, PROLONG starts from the ground up with observed national deployment records. The framework feeds historical wind and solar installation data from dozens of countries into a Monte Carlo engine that generates thousands of plausible “virtual worlds,” each representing a different trajectory a nation might have followed given its policy environment, grid conditions, and market dynamics. A machine-learning layer then trains on those synthetic histories to produce probability distributions for future capacity, not single-point forecasts.
That probabilistic design is one of the study’s central contributions. Instead of offering a lone headline number, PROLONG gives policymakers a spread: a median expectation flanked by bands showing how likely faster or slower buildout paths are. The authors explicitly benchmark their outputs against two major reference points from the Intergovernmental Panel on Climate Change: the scenario ensembles in the IPCC’s Sixth Assessment Report and the 1.5-degree-consistent pathways from the 2018 Special Report on Global Warming of 1.5°C.
Within the time horizons where comparison is possible, PROLONG’s median projections tend to match or exceed the upper range of those conventional scenarios. The implication is stark: deployment momentum already embedded in national track records may be outpacing the assumptions baked into international climate targets.
Open data strengthens the case
One reason the study has drawn attention quickly is transparency. The authors released source data files covering every figure and extended data chart, including the spreadsheets behind their hindcast accuracy tests. Those tests replay the model against historical periods it was not trained on, checking whether PROLONG can track past growth trajectories with meaningful precision.
Because the data are public, any researcher or analyst can reproduce the key results, inspect uncertainty intervals, and stress-test the model’s performance claims without relying on third-party summaries. That open evidence trail sets PROLONG apart from proprietary forecasting tools whose internal mechanics remain opaque.
The IEA’s Renewables 2025 report provides a useful external benchmark. The agency’s regularly updated outlook is among the most widely cited near-term projections from an intergovernmental body, and its repeated upward revisions have become a running theme in energy analysis. PROLONG’s authors use the IEA outlook as one of several comparison points, testing whether their simulation-driven approach captures the same acceleration the agency has been forced to acknowledge through successive forecast corrections.
Where the model faces pushback
PROLONG’s core strength, learning from historical national deployment curves, is also its most obvious vulnerability. Countries that built renewable capacity rapidly often did so under specific policy regimes, supply-chain conditions, and grid architectures that may not transfer cleanly to nations at earlier stages of the transition. Grid bottlenecks, permitting delays, and political reversals are difficult to capture in a framework that primarily extrapolates from past installation rates.
The study’s own uncertainty bands acknowledge these risks, but their reliability at longer time horizons has not yet been independently validated. Because the model trains on the historical period for which consistent data exist, projections beyond the next decade or two involve extrapolation into policy and technology environments that could diverge sharply from the past. Structural breaks, such as major shifts in global trade, abrupt changes in public support, or unforeseen technical constraints, are hard for any data-driven model to anticipate.
National policy landscapes have also shifted since much of the training data was compiled. Legislation in the United States, the European Union, and China has redrawn the incentive map for renewable investment, while trade disputes over solar panel supply chains have introduced new friction. Whether PROLONG’s simulations, built from pre-existing deployment patterns, adequately account for these recent policy shocks is a question the published paper does not fully resolve.
Another concern is granularity. PROLONG works with aggregated country-level data, which can smooth over local constraints in transmission build-out, land-use conflicts, and regional differences in resource quality. As wind and solar penetration climbs toward very high shares of the electricity mix, those subnational bottlenecks could become more binding than national averages suggest.
Why the gap with legacy models matters for climate planning
No official statement from the IPCC addresses whether PROLONG’s comparison to its scenario ensembles is methodologically sound. The IPCC pathways were built for a different purpose: mapping emissions trajectories under various policy and technology assumptions. Comparing them directly to a deployment-pattern model involves framing choices that reasonable analysts could dispute. External modeling groups have not yet published systematic comparisons between PROLONG and their own tools, so the broader research community’s verdict is still forming.
For policymakers, the practical takeaway is narrower but significant. If the historical record of national renewable buildouts already implies faster growth than legacy models assume, then planning frameworks anchored to those older baselines may be systematically underestimating what is achievable. That has consequences for grid investment timelines, fossil-fuel phase-out schedules, and the credibility of nationally determined contributions under the Paris Agreement.
The most defensible reading of the evidence as of May 2026 is that PROLONG offers a credible, data-rich challenge to conventional wisdom about renewable deployment, backed by transparent methods and accessible source files. Its projections should be treated as provisional, not definitive. As other researchers engage with the released data, replicate the results, and test alternative model specifications, the field will gain a clearer picture of how much weight this new framework deserves in the broader landscape of climate and energy modeling.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
