Blog
ARISE Aims to Cut Geothermal Exploration Costs
Estimated reading time: 4 minutes · Last updated: 2026-08-20
Portland State University will lead ARISE, a U.S. Department of Energy–supported national team that applies artificial intelligence to make geothermal exploration less risky and cheaper. The project combines Stanford models, a U.S. Geological Survey machine‑learning team and PSU's ARID algorithm to zone geologically similar areas and target the most valuable next measurements. In a nine‑month first phase the team aims to reduce the uncertainty in exploration cost estimates by at least 10 percent on average, testing its recommendations against the measurement history from the DOE‑funded Utah FORGE research site.
“We use AI and years of historical data to make that bet less of a gamble.”
John Lipor, Wedge Vision Associate Professor of electrical and computer engineering at PSU
Key takeaways
- Portland State University will lead ARISE, a DOE‑supported collaboration that includes Stanford University, the U.S. Geological Survey and 400C Energy.
- The nine‑month first phase aims to reduce cost‑estimate uncertainty by at least 10 percent on average compared with current methods.
- Deliverables include a new underground temperature map for Oregon and public release of models, data and code through DOE's Geothermal Data Repository.
- DOE analysis projects enhanced geothermal could expand from close to 4 gigawatts today to between 90 and 300 gigawatts by 2050, if exploration costs fall.
Table of contents
How ARISE targets the exploration gamble
Exploration for geothermal power is expensive because true temperatures miles underground can only be confirmed by drilling. ARISE approaches that problem by combining three existing tools from the project partners: country‑scale temperature prediction from Stanford, a price‑impact module that translates temperature uncertainty into dollar terms, and PSU's own algorithm that divides the map into more uniform geological zones so models do not have to cover the entire country at once.
PSU's algorithm reduces the need for a single nationwide model by giving each zone its own tailored predictor. That makes local temperature estimates sharper and lets the system prioritise where a single new measurement will reduce the most cost uncertainty. In plain terms: rather than treating the U.S. as one climate, ARISE treats each geological type separately and targets measurements that shrink the developer's financial risk.
The nine‑month test and how success will be judged
The project begins with a defined nine‑month first phase that carries an explicit numerical target. The team will replay the measurement history from the DOE‑funded Utah FORGE site and compare what ARISE would have recommended with the actual choices engineers made on the ground. The stated performance goal for this phase is to reduce the uncertainty in cost estimates by at least 10 percent on average versus the method now in use.
That comparison converts geological uncertainty into dollars using Stanford's module that maps temperature ranges into likely electricity‑price outcomes, so the evaluation is financial as well as geological. If the AI recommendations consistently pick measurements that cut the cost range faster than current practice, the team will treat the phase as a success and plan to package the workflow for developer use and to test it in other regions.
Why cheaper exploration matters for baseload power
Geothermal is a round‑the‑clock, carbon‑free power source that can meet steady baseload demand. DOE analysis cited by the project team estimates enhanced geothermal systems could grow installed capacity from close to 4 gigawatts today to between 90 and 300 gigawatts by 2050 if exploration costs fall enough to unlock widespread development.
The project also frames the need in terms of rising steady demand: global electricity use by data centres is projected to more than double by 2030, and U.S. data centres alone could account for up to 12 percent of national demand. Lowering the cost and risk of finding high‑temperature resources is therefore central to whether geothermal can scale to meet that kind of continuous load.
| Component | Provider | Role |
|---|---|---|
| Temperature prediction model | Stanford University | Produces country‑scale temperature estimates |
| Price‑impact module | Stanford University | Translates temperature uncertainty into electricity‑price outcomes |
| Regionalisation algorithm (ARID) | Portland State University | Groups similar geology to let zone‑specific models improve precision |
| Field test data | Utah FORGE (DOE‑funded) | Replay history used to evaluate ARISE recommendations |
The case for and against ARISE scaling geothermal exploration
The case for
- If ARISE meets its nine‑month 10 percent target, developers could make fewer, better‑timed measurements and cut upfront exploration costs, accelerating project pipelines.
- Open release of the team's models, data and code through DOE's Geothermal Data Repository would let other researchers and companies validate and build on the approach quickly.
The case against
- Models trained on historical records may underperform in geologically atypical areas, limiting gains outside the regions represented in training data such as Utah.
- Even a successful tool reduces only exploration uncertainty; permitting, grid access and drilling supply constraints would still restrict how fast new capacity reaches operation.
What to be careful about
- ARISE's performance target applies to averages; some zones could see smaller or no improvement, leaving developers exposed where gains are weakest.
- The Utah FORGE replay tests historical decisions, which may not capture future commercial constraints such as site access, drilling lead times or local permitting changes.
The bottom line
ARISE packages existing temperature and price models with a PSU regionalisation algorithm and a measurement‑selection step into a testable workflow designed to lower the financial risk of drilling for geothermal heat. The nine‑month phase centred on the Utah FORGE replay sets a clear numerical bar — at least a 10 percent average cut in cost‑estimate uncertainty — and the team intends to publish maps and code through DOE channels if the test succeeds. Those outputs will determine whether ARISE becomes a practical tool developers can use to speed geothermal deployment at scale.
What to watch
- Watch for the team’s report at the end of the nine‑month phase; no date has been set.
- Watch for the public release of the Oregon underground temperature map and the project's models and code to DOE's Geothermal Data Repository; no date has been set.
- Watch for plans to test the system at the Newberry volcanic area in central Oregon; no date has been set.
Frequently asked questions
What is ARISE and who is leading it?
ARISE is an AI‑based effort to regionalise temperature prediction and prioritise measurements for enhanced geothermal systems; Portland State University leads the multi‑institution team that includes Stanford University, the U.S. Geological Survey and 400C Energy, and is supported by the U.S. Department of Energy.
How will the project measure success in its first phase?
The nine‑month first phase will be judged by whether the system can reduce uncertainty in exploration cost estimates by at least 10 percent on average compared with current methods, using a replay of measurement records from the DOE‑funded Utah FORGE research site.
When will the team's models and data be available?
The team plans to release models, data and code through DOE's Geothermal Data Repository; the material states that this is a planned deliverable but gives no public release date.
Related reading