A second life for fossil-fuel infrastructure
Old oil wells, depleted reservoirs and the skills that built them could carry clean, firm power. Where, how fast, and at what cost?
Every number here comes from published SEAR Lab papers, listed at the bottom with the table or section each one comes from. Where a paper has an error, the demo shows the corrected value and says so. Cite the papers, not this page.
1. Hot rock under old oil country
Eight states from Nevada to Mississippi hold thousands of orphaned oil and gas wells, many on hot ground, on tribal lands or in HUBZones (federally designated low-income business areas). Pick a finding to see which states the paper names for it, then break the potential down by how hot the rock is.
By geothermal class
Correction. The paper's Conclusions give 2,328 GW of technical potential and 3 GW from converted wells; its Table 4 and Section 4.8 give 2,372.8 GW and 4.3 GW. The demo uses Table 4. (Table 4's class rows are rounded, so they add to 2,372.9 GW and 4.2 GW.) The Conclusions' other figures agree with Table 3: 372,309 sq mi at Class 1–3, and 6,166 orphaned wells in HUBZones on Class 1–3 ground.
The paper publishes no numbers by state, so the map shows only which states it names for each finding. Potential is counted only on ground at 120 °C or more (Classes 1–3); Classes 4–5 appear only in land area and well counts. The totals are a technical upper bound: land use, economics and site checks are left for future work.
2. CO₂ pipelines: build it all at once, or in five-year phases?
Carbon capture sends CO₂ by pipeline to deep saline rock or to oil fields for enhanced recovery. In a Gulf-region case (80 sources, 125 storage sites, 30 years), the same project was planned as one design and as six five-year phases. Step through the phases and watch where the money goes.
"Cost so far" multiplies each phase's published cost per year by the years elapsed, so it is exact at every year shown. Negative costs in the paper's tables are net revenue from storage credits; the demo shows transport, where the two designs differ.
Check. In Table 1 the single-phase annual cost (−$34.8M/yr) does not equal its rounded parts ($126M − $180M + $18.5M = −$35.5M/yr); the phased columns agree with their parts to within about $0.2M. The transport rows, which this panel uses, add up exactly to the printed totals ($555M and $363.65M). In "only what pays" mode the paper's "5% more CO₂" is 3.03 against 2.87 Mt/yr, which is 5.6%.
3. Geothermal as firm power for an AI data center
A 250 MW data center in Texas can build its own power on site while staying on the grid. Six mixes were scored on blended cost and on how much of the grid's outage risk they remove. Pick a mix; compare the two geothermal-anchored designs, S3 and S4.
Installed at the site (Table 11) and what it can carry alone in an outage (Table 15)
Correction. The paper's abstract says gas-dependent mixes range from 58.0% to 91.2% outage protection. In its Table 15 the 58.0% belongs to S4 (geothermal + solar + grid), which needs no fuel; the gas-dependent mixes (S1, S2, S6) run from 72.5% to 91.2%.
Outage protection is the paper's avoided loss-of-load probability; it depends on the paper's stated assumptions about outages, battery bridging and fuel supply, and is not a guarantee. S3's ">99.9%" is drawn at 99.9%. S3's geothermal is 250 MW in Table 11 but 300 MW of nameplate in Table 13; the demo uses Table 11. A solar PPA is a contract for utility-scale solar power, not panels on site.
What this shows
The oil and gas era left behind more than emissions. It left drilled wells, mapped hot rock, pipeline corridors, depleted reservoirs and a workforce that knows how to drill. Each of these papers asks how some of that could be put back to work.
The first panel is about place. The ground that is hot enough for enhanced geothermal plants overlaps with orphaned wells, tribal lands and low-income business zones, so federal programs for those communities could steer the benefits. Switch the land filter to "Tribal": about 15% of the technical potential lies on tribal land, but on Class 3 ground tribal land holds a quarter of the potential and nearly half of the converted-well capacity.
The second panel is about timing. Building a CO₂ network all at once means paying to run its longest pipelines from the first year. Planning in phases starts with short links between nearby sources and storage, and adds the long, costly routes only when the close storage fills. Drag the phase slider: the phased plan spends more per year than the all-at-once design only in its last phase, and when both must store 3 Mt a year it ends the 30 years about a third cheaper on transport.
The third panel is about firmness. A data center needs power every hour. Geothermal runs around the clock with no fuel pipeline, so an all-geothermal site (S3) gives the strongest outage protection at a moderate cost. The cheapest mix (S4) also uses geothermal, but its on-site plants can carry only 145 of the 250 MW, so it protects least. Cheapest and most resilient are not the same design.
Behind this demo
Sources
- Jones, E.C., Jr., Munjurpet Sridharan, C., Aghapour, R., & Rodriguez, A. (2025). Re-energizing legacy fossil infrastructure: Evaluating geothermal power in tribal lands and HUBZones. Sustainability, 17(6), 2558. doi:10.3390/su17062558. Panel 1: Tables 1, 3, 4 and 5; Sections 4.1, 4.3, 4.5, 4.6, 4.8, 5 and 6. CC BY 4.0.
- Jones, E.C., Jr., Yaw, S., Bennett, J.A., Ogland-Hand, J.D., Strahan, C., & Middleton, R.S. (2022). Designing multi-phased CO₂ capture and storage infrastructure deployments. Renewable and Sustainable Energy Transition, 2, 100023. doi:10.1016/j.rset.2022.100023. Panel 2: Tables 1 and 2; Sections 4.1 and 4.2. Numbers only; the paper's figures are not reproduced (CC BY-NC-ND 4.0).
- Jones, E.C., Jr., & Jones, E.C., Sr. (2026). Megawatts to zettaflops: A techno-economic framework for grid-tied behind-the-meter architectures in AI data centers. Electricity, 7(2), 43. doi:10.3390/electricity7020043. Panel 3: Tables 8, 11, 14 and 15. CC BY 4.0.
- State outlines: U.S. Census Bureau, 2023 cartographic boundary file (
cb_2023_us_state_20m); projected to Albers Equal Area and simplified offline (_build/build_states.py). No basemap tiles.