The U.S. Geological Survey has published a provisional assessment estimating roughly 135 gigawatts of electric generation capacity locked beneath the Great Basin in the southwestern United States, enough to supply about 10 percent of the nation’s electricity demand. Reaching that potential depends on advances in enhanced geothermal systems, or EGS, and the federal agencies driving the work are turning to artificial intelligence and machine learning to identify where to drill, how to model underground heat, and how to speed up permitting for new projects.
135 Gigawatts Beneath the Great Basin
The scale of the resource is striking. A recent USGS assessment on the EGS electric-resource potential for the Great Basin places the provisional estimate at approximately 135 GWe of generation capacity drawn from the upper 6 km of the Earth’s crust. That figure is conditional on technology advances, meaning it represents what could be extracted if drilling techniques, reservoir engineering, and heat-exchange systems continue to improve at their current pace.
To put 135 GWe in perspective, the entire U.S. nuclear fleet operates at a comparable scale of capacity. If the Great Basin’s geothermal reserves were fully developed, they would represent a baseload energy source that runs around the clock, unlike solar and wind installations that depend on weather. A corresponding USGS release tied to the assessment frames the opportunity in blunt terms: enhanced geothermal systems in the Great Basin could supply 10 percent of U.S. electricity demand.
That 10 percent figure is not a forecast but a technical potential, contingent on both engineering progress and sustained investment. It assumes the industry can routinely drill to several kilometers, fracture hot but relatively dry rock, and circulate working fluids through engineered reservoirs without unacceptable seismic or environmental impacts. Even if only a fraction of the 135 GWe is ultimately realized, it would still represent one of the largest additions of firm, low-carbon capacity in U.S. history.
Machine Learning Sharpens the Search for Heat
Finding the right place to drill has always been geothermal energy’s most expensive gamble. Exploratory wells can cost millions of dollars, and a dry hole yields nothing. That risk calculus is what makes AI-driven site selection so consequential. USGS scientists working under the INGENIOUS initiative have built supervised machine learning models that ingest geological, geophysical, and geochemical data to produce hydrothermal favorability maps for the Great Basin, replacing much of the manual pattern-recognition work that geologists once performed by hand.
The practical value is straightforward: instead of relying on surface-level surveys alone, researchers can now rank thousands of potential drill sites by probability of success before a single rig is mobilized. The models assign prediction percentiles to each location, giving project developers a data-driven way to prioritize where they spend capital. That shift from intuition-heavy exploration to algorithmic targeting could compress the timeline from initial survey to productive well by years, and it offers a way to systematically update prospects as new subsurface data comes in.
USGS leadership has explicitly credited artificial intelligence and machine learning with speeding up and improving the accuracy of broader geothermal assessments, situating the work within the framework of the Energy Act and related mandates. The agency’s public materials emphasize that these tools are not experimental add-ons but are becoming embedded in how USGS programs evaluate subsurface energy resources, from data integration and anomaly detection to probabilistic resource estimation.
DOE Funding and the Broader AI Push
The Department of Energy has been building its own AI toolkit in parallel. In July 2021, the agency selected four projects to receive up to $3.5 million to advance research in machine learning for geothermal energy. Those awards targeted the gap between raw geological data and actionable drilling decisions, funding teams that apply algorithms to reservoir characterization, temperature prediction at depth, and real-time optimization of field operations.
DOE’s geothermal office has also highlighted how advanced algorithms could better identify areas for exploratory drilling and drive operational improvements across the project lifecycle. In its dedicated machine learning overview, the agency points to use cases such as mapping hidden geothermal systems, forecasting reservoir performance, and automating fault detection in production data. The language is forward-looking, acknowledging that the technology is still maturing, but the investment pattern is clear: federal planners expect predictive models to lower the financial risk that has historically kept geothermal development slower than wind or solar.
That AI push extends beyond subsurface science. DOE notes that AI tools for energy are being developed to improve how projects are sited and permitted at the federal, state, and local levels, with the goal of supporting faster deployment of new infrastructure. Permitting delays are among the most cited obstacles to new geothermal capacity. If algorithms can pre-screen parcels for environmental conflicts, flag data gaps in applications, and standardize portions of environmental review, they could shave months off project timelines and reduce soft costs for developers.
Data policy is another piece of the puzzle. DOE requires that recipients of geothermal research funding follow specific data provision instructions, ensuring that subsurface measurements, modeling results, and operational datasets flow into shared repositories. That growing pool of standardized data is essential for training and validating the next generation of machine learning models, which typically perform better when exposed to diverse examples from multiple geologic settings.
Why Geothermal Still Lags and What AI Changes
Most coverage of geothermal energy treats it as a sleeping giant, and for good reason. Despite decades of research, geothermal supplies less than one percent of U.S. electricity. The main barriers are well understood: high upfront drilling costs, geological uncertainty, and a permitting process that was designed for oil and gas wells rather than heat-mining operations. AI does not eliminate any of those barriers outright, but it attacks the uncertainty problem directly and, in doing so, can indirectly reduce both cost and delay.
Consider the difference between conventional geothermal development and the AI-assisted approach now emerging from USGS and DOE research. In the traditional model, geologists identify a promising area based on surface indicators like hot springs or volcanic activity, then drill test wells that may or may not confirm a viable reservoir. Each failed well burns through capital and time, and even successful fields can underperform if fractures and temperature gradients are not well understood.
In the AI-assisted model, machine learning processes thousands of data points (gravity anomalies, magnetotelluric surveys, fault maps, well logs, and geochemical signatures) to estimate the likelihood of encountering hot, permeable rock at depth. Instead of a handful of hand-drawn prospects, developers can work from ranked portfolios of candidate sites, each with quantified uncertainty ranges. During development, similar tools can help design stimulation plans for EGS reservoirs, monitor microseismicity, and adjust injection strategies based on real-time feedback.
Permitting stands to change as well. Environmental reviews for geothermal projects often require extensive analysis of groundwater impacts, induced seismicity, wildlife habitat, and cultural resources. AI cannot replace human judgment in these domains, but it can sift through large document archives, geographic information system layers, and historical case files to highlight relevant precedents and potential conflicts early in the process. That, in turn, can help agencies and developers focus field studies where they matter most and avoid late-stage surprises.
None of this guarantees that the Great Basin’s 135 GWe of potential will be realized. Technical challenges remain in drilling deep, managing induced seismicity, and ensuring that EGS reservoirs remain productive over decades. Community acceptance and careful environmental stewardship will be essential, particularly in regions where geothermal development overlaps with sensitive ecosystems or cultural sites.
What the USGS assessment and DOE initiatives do show is that the underlying physics of geothermal energy (abundant heat at depth and the ability to tap it with engineered systems) is increasingly paired with digital tools that can make the resource more predictable and financeable. If AI and machine learning can consistently reduce the number of dry holes, improve reservoir performance, and streamline permitting, they could turn the Great Basin’s theoretical endowment into a cornerstone of a more resilient, low-carbon grid.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
