Why can't we reliably create the right fracture network?
The core problem is that hydraulic stimulation — injecting fluid to crack hot rock — often produces fractures that are either too few, too uneven, or too short-lived to sustain efficient heat extraction. One review of 41 EGS projects worldwide found that the fracture network and reservoir quality induced by stimulation are essentially uncontrollable, leading to insufficient reservoir volume, unstable fractures, severe fluid loss, and unacceptable induced earthquakes [5]. This unpredictability is the fundamental reason EGS has not achieved commercial scale, with cumulative installed power generation reaching only 37.41 MW by the end of 2021 — a tiny fraction of what would be needed for commercial viability [5].
A modeling study of the Basal Cambrian Sandstone in Alberta, Canada, showed that while hydraulic fracturing can accelerate early energy harvesting, the benefit shrinks at higher injection rates, and the energy produced versus energy invested ratio ranges from 4 to 9 depending on operating conditions [1]. This wide range reflects the difficulty of predicting how a given stimulation will perform. The study also found that the greater the injection rate, the smaller the benefit of fracturing, suggesting that simply pumping harder is not a solution [1].
Can we predict and control induced earthquakes?
Induced seismicity is a major barrier to public acceptance and regulatory approval of EGS, but current forecasting tools are not reliable enough. A 2025 study tested machine learning models on data from three EGS sites — Cooper Basin (Australia), St1 Helsinki (Finland), and a laboratory experiment — and found that while feature-rich models performed well in complex seismic environments like Cooper Basin, they struggled with two critical issues: data scarcity from operational wells and the difficulty of predicting sudden jumps in seismic moment for large events [2]. In simpler settings like St1 Helsinki, where events were not clustered, adding more features did not improve predictions [2].
The authors concluded that reliable near-real-time forecasting would require synthetic data augmentation and better feature selection to capture diverse reservoir dynamics [2]. Without these advances, operators cannot confidently adjust injection in real time to avoid triggering a damaging earthquake, which keeps EGS projects in a high-risk category for regulators and insurers.
How do we stop cold water from shortcutting through the reservoir?
Early thermal breakthrough — where injected cold water finds a preferential path and returns to the production well before absorbing enough heat — is a persistent economic killer for EGS. A 2023 study proposed a novel solution: proppants (small particles that prop open fractures) that expand or contract with temperature to autonomously adjust fracture conductivity. In field-scale simulations, these smart proppants increased heat extraction by 31.4% over 50 years compared to conventional proppants, effectively delaying thermal breakthrough and sweeping heat from a larger rock volume [3].
However, this technology is still at the conceptual and microscale modeling stage; the study explicitly notes that the required material properties (negative thermal expansion coefficients) need to be developed and tested in real reservoirs [3]. Another study from 2021 showed that tracer tests combined with data assimilation can help characterize fracture flow patterns and predict thermal performance, but the method still requires extensive data and careful uncertainty handling [4]. Together, these studies indicate that while promising approaches exist, we lack field-validated tools to prevent thermal breakthrough — a gap that directly threatens the economic viability of any EGS project.
Why can't we just copy a successful EGS design from one site to another?
The most fundamental evidence gap is the absence of a reproducible thermal reservoir stimulation model. A comprehensive review of 41 EGS projects concluded that the dependence of stimulation technologies on in-situ geological conditions — rock type, stress state, natural fractures, temperature — means that a technique that works at one site often fails at another [5]. The review calls for establishing a global database of hot dry rock and EGS characteristics to build an accurate quantitative system linking geological conditions to reservoir reconstruction outcomes [5].
This lack of transferability is why, after more than 50 years of research and 41 projects, EGS remains on the learning curve rather than at commercial scale [5]. Until we can predict how a given stimulation method will perform in a new geological setting, each EGS project is essentially a one-off experiment, which makes investment risky and slows deployment.
About These Sources
This answer is built on 5 peer-reviewed studies — published from 2021 to 2025, 2 from 2024 or later, 1 in Q1 journals, collectively cited 82 times — selected as the most relevant from 5 studies that passed quality screening, drawn from 67 papers retrieved from a database of over 500 million.
Sources used in this answer
Investigation of enhanced geothermal system in the Basal Cambrian Sandstone Unit, Alberta, Canada
Modeling of the Basal Cambrian Sandstone in Alberta, Canada, shows that hydraulic fracturing accelerates early energy harvesting but the benefit shrinks at higher injection rates, with energy produced-to-invested ratios ranging from 4 to 9 depending on operating rate [1].
Forecasting induced seismicity in enhanced geothermal systems using machine learning: challenges and opportunities
Machine learning models for forecasting induced seismicity in EGS (tested on Cooper Basin, St1 Helsinki, and lab data) struggle with data scarcity and predicting large seismic events; feature-rich models help in complex settings but not in simpler ones [2].
Autonomous Fracture Conductivity Using Expandable Proppants in Enhanced Geothermal Systems
A novel concept using expandable proppants with negative thermal expansion coefficients could autonomously control fracture conductivity, potentially increasing heat extraction by 31.4% over 50 years in field-scale simulations, but remains at the conceptual stage [3].
Predicting Thermal Performance of an Enhanced Geothermal System From Tracer Tests in a Data Assimilation Framework
A data assimilation framework using tracer tests can characterize fracture flow and predict thermal performance in EGS, but requires extensive data and careful handling of uncertainties [4].
Challenges and opportunities of enhanced geothermal systems: A review
A review of 41 EGS projects worldwide (cumulative installed capacity only 37.41 MW by end of 2021) identifies the lack of a reproducible stimulation model due to dependence on site-specific geological conditions as the fundamental barrier to commercialization [5].