Introduction

Wind and solar photovoltaics (PV) have experienced dramatic cost reductions and are poised to play a central role in future low-carbon electricity systems.1 Both technologies produce fluctuating renewable energy, requiring complementary storage technologies2 but also “firm” clean energy sources which provide flexible and controllable supply.3 One such technology is geothermal power, which utilizes subsurface heat and offers renewable firm electricity on a global scale.4,5 Despite its long-standing presence, geothermal power deployment has been slow, and its application remains limited to a few, mostly volcanic, regions.4 One reason is the high cost of drilling in regions where the heat source sits deeply underground.4,5,6,7

Recently, interest in geothermal power has risen sharply, driven by great hopes that its costs will fall substantially.4,8 However, while multiple studies on the cost of geothermal power exist,9,10,11,12 their cost dynamics remain poorly understood.8 Reasons for this are the scarcity and quality of long-term data and local heterogeneity of the technology and its use environment, and are discussed in more detail in the following paragraph. This lack of understanding is in sharp contrast to other renewable energy technologies, where cost dynamics have been analyzed in depth and found to be a decisive factor in their competitiveness vis-à-vis each other but also fossil fuel-based technologies.13,14,15 An important and well-established concept in this regard is the experience curve (often also called learning curve).14,16,17 It describes the empirical phenomenon that the relative cost of a technology is reduced by a relatively constant percentage—the so-called experience rate (ER)—per doubling of cumulatively installed capacity.13 Experience curves are arguably the most reliable method to assess technological cost trajectories.18 The two existing global studies on geothermal experience curves report strongly negative ERs, between negative 29% and negative 100%.19,20 These numbers stand in sharp contrast to the International Energy Agency’s 2024 geothermal report, which uses expert elicitation to project cost decreases of up to 80% by 2035,4 implying extremely high ERs even under optimistic deployment scenarios.

The existing analyses, of which we provide an overview in Table S1, are plagued by several issues related to data: (1) There are too little data to construct reliable experience curves. The literature recommends considering data over at least three doublings of capacity,21 but the dataset from IRENA that many authors rely on only covers 0.3 doublings of capacity. Also, its data only starts from 2008, providing little insight into true long-term trends. (2) The data used lacks proper verification, and our analysis has found several flaws in the data. This leads to incorrect results, and single errors have a major impact due to the small sample size. More generally, the lack of high-quality data has been identified as a major challenge in constructing geothermal experience curves.13,22 (3) The analyses ignore local differences in the cost dynamics. Geothermal power can be expected to have locally differing costs due to a high local cost share and differences in plant design. Ignoring these differences can skew results. For example, the analyses might interpret a shift to higher-cost countries with negative learning. The analyses provide little explanation for the observed experience rates. Without a differentiated analysis of ER drivers, it is hard to reason if and how new technologies or policies can change the past ER.16,17

The hopes for cost improvements are partly based on two emerging geothermal technologies: enhanced geothermal systems (EGS) and closed-loop geothermal systems (CLGS), sometimes also called advanced geothermal systems. These technologies promise greater scalability than the existing hydrothermal technology. Hydrothermal systems rely on rock permeability and the existence of water in the underground, both of which can vary strongly locally and are costly to assess.23 EGS uses recently developed fracking technology to make plants independent from natural—often too low—permeability.24 CLGS utilize a closed underground heat exchanger with many thin wells and circulating a selected fluid.25,26,27 Despite the apparent advantages of EGS and CLGS, it remains unclear whether these technologies’ characteristics truly result in substantially lower long-term cost. Ultimately, geothermal power needs realistic long-term improvement in technological learning to become competitive globally.

In this study, we address the uncertainty around geothermal cost dynamics by providing the first robust empirical estimate of global ERs, drawing on a newly compiled dataset covering all 514 geothermal power plants that have been commissioned over the last seven decades. To understand the large variation in reported costs across geographies, we also estimate local, i.e., country-specific ERs. Because experience curves quantify the rate of cost change but not its underlying causes, we complement this quantitative analysis with 51 interviews with geothermal experts across the world, identifying key drivers of and barriers to cost reductions. This mixed-methods design, established in prior learning studies of solar PV,28,29 wind,29 nuclear power,30 and financing cost of renewable energy in general,31 lets us link the observed ER patterns to their structural determinants. Finally, we address the future technologies EGS and CLGS. Drawing on a recently established framework of technology characteristics and expert ratings,32,33,34 we assess the potential for these emerging technologies to enable faster and global cost declines. Together, these steps provide a comprehensive understanding of past and future geothermal cost dynamics and their implications for energy policy and modeling. Our contribution is primarily empirical: We provide the first consistent global learning rate estimate for geothermal power based on a comprehensive, multi-decade dataset and offer directly actionable insights for energy system modelers seeking to understand where and under what conditions geothermal can become cost-competitive.