Power Without End: Geothermal Co-Location Changes Everything
Weakens the Case for the Powerlink
The prior analysis led with natural gas as the "honest answer" for AI data center baseload power. That was wrong for Imperial Valley specifically. The honest answer there is geothermal co-location — a model that is actively being built, requires no grid connection, bypasses CAISO entirely, and in doing so significantly undermines one of the primary justifications for the Golden Pacific Powerlink.
The prior version of this analysis opened with a declaration: AI data centers "cannot run on solar panels," and the honest near-term power source was natural gas. That remains true globally, and for most of the United States. But Imperial Valley, California, is not most of the United States. It sits atop the Salton Sea Geothermal Field, one of the largest proven geothermal resources in the world, with estimated developable capacity of 2,250 megawatts — of which only about 403 MW is currently in operation. Crucially, it is also served by the Imperial Irrigation District, a publicly owned utility that is entirely independent of CAISO and the state's investor-owned utility grid.
These two facts, taken together, produce a clean energy architecture for AI data centers in Imperial Valley that does not require natural gas, does not require the CAISO grid, and does not require the Golden Pacific Powerlink. The data centers generate their own power on-site from geothermal energy, consume it behind the meter, and the electrons never touch the SDG&E transmission system. This is not a theoretical future state. It is being actively built by multiple developers, with the explicit intent of serving AI computing workloads.
The analytical implication is significant: to the extent that the Golden Pacific Powerlink's need case depends on AI data center load growth in Imperial Valley appearing on the CAISO grid, the co-location model removes that load from the equation. The Powerlink's justification narrows — to grid reliability and renewable integration arguments that are real, but materially more modest than the combined need case SDG&E and CAISO have presented.
The Corrected Power Source Picture for Imperial Valley
The previous power source table placed natural gas first, geothermal as a footnote, and characterized California as facing a 24/7 baseload "contradiction" that had no clean near-term solution. The corrected table, specific to Imperial Valley, looks fundamentally different.
| # | Source | 24/7 Firm | Imperial Valley Status | Co-Location Viable? | Powerlink Needed? | Grade |
|---|---|---|---|---|---|---|
| 1 | Geothermal (conventional hydrothermal) | ✅ 90–98% CF | 403 MW operating; 1,850 MW developable proven resource with 40-year operating history | ✅ Yes — actively being built at CTR, CalEthos | ❌ No — behind-the-meter on IID grid | A |
| 2 | Solar + geothermal hybrid | ✅ Effectively 24/7 with geo baseload | Abundant solar; geothermal provides evening/overnight firm power | ✅ Yes — CalEthos model (420 MW, 100% geothermal + solar) | ❌ No — behind-the-meter | A |
| 3 | Natural gas (on-site or grid) | ✅ Fully dispatchable | Available; contradicts California climate policy; IVCM acknowledges gas backup | ⚠️ As backup only; ESG liability | ⚠️ Reduces need but doesn't eliminate it for grid-connected projects | C |
| 4 | Nuclear — existing large-scale | ✅ Ideal baseload | None in California except Diablo Canyon; no new projects planned | ❌ Not site-flexible | N/A | D for CA timeline |
| 5 | SMR — Small Modular Reactors | ✅ When available | No California projects; post-2030 at earliest nationally | ⚠️ Eventually; not now | ❌ No — would be co-located | C for timeline |
| 6 | Solar only (no storage) | ❌ ~6 hrs/day | Extremely abundant but cannot provide 24/7 AI workload power alone | ❌ Insufficient alone | N/A | D alone |
What Co-Location Actually Means
The term "co-location" in this context describes a specific power delivery architecture. The geothermal plant and the data center occupy the same campus, or are connected by a short, dedicated power line. The data center's electricity never enters the public grid — it flows directly from generator to load, entirely behind the meter. The data center operator either owns the geothermal plant, has a long-term power purchase agreement with the geothermal developer, or both.
⚡ Geothermal Plant
Salton Sea field
90–98% capacity factor
24/7 firm baseload
No weather dependence
🖥 AI Data Center
Behind-the-meter
No CAISO connection
No SDG&E transmission
No Powerlink required
The Rhodium Group quantified this opportunity in a 2025 report: 55 to 64% of the projected growth in hyperscale data center capacity could be met with behind-the-meter geothermal power, representing 15 to 17 gigawatts of new capacity. In the scenario where data center siting decisions follow geothermal resources rather than historical fiber-network clustering near cities, geothermal could meet all projected data center load growth by the early 2030s.
By placing data centers adjacent to geothermal resources, operators can minimize transmission losses, avoid exposure to lengthy interconnection queues, reduce infrastructure build-out costs, and secure long-term energy price stability. This approach enables hyperscale facilities to scale alongside geothermal development.
The IID Advantage: Bypassing CAISO Entirely
The Imperial Irrigation District is one of the few large utilities in California that is not a CAISO participant. It is a publicly owned utility serving Imperial County and parts of Riverside County, operating its own transmission and distribution system under its own governance. This institutional independence has a cascade of practical consequences that are not widely understood outside the industry.
A data center connecting to IID's grid avoids: CAISO's generator interconnection queue (currently running 4 to 7 years in major markets); FERC formula rate cost allocation that socializes transmission costs across all CAISO customers; SDG&E's transmission owner rate structure; and the entire Powerlink cost-sharing mechanism. It also avoids California's most complex utility regulatory processes — CPUC proceedings, CPUC rate cases, and the multi-year Powerlink approval process itself.
The practical result, as one local analysis of the Imperial Valley data center opportunity put it: "IID's independence from CAISO means that a developer connecting to IID's system avoids the interconnection queue and transmission constraints that have delayed comparable projects in CAISO territory for years. The straightforward regulatory environment — one utility, manageable interconnection process, locally-controlled decision-making — is a competitive advantage that shortens project timelines significantly." Transmission upgrades required to support large new loads on IID's system are managed locally, without FERC cost socialization. The ratepayer protection problem that afflicts CAISO-connected data centers does not arise in the same way for IID-connected campuses.
If AI data centers in Imperial Valley connect to IID's grid rather than SDG&E's, and generate their own power from co-located geothermal, then: (1) the data center load never appears on the CAISO system; (2) the Powerlink carries no data-center-driven justification; and (3) the cost socialization mechanism that would charge San Diego residential ratepayers for data center infrastructure does not apply. The co-location model simultaneously solves the power source problem, bypasses the ratepayer equity problem, and weakens the demand case for the Powerlink — all at once.
What Is Actually Being Built Right Now
This is not a theoretical architecture. Multiple developers are actively pursuing it in Imperial Valley.
Controlled Thermal Resources (CTR) — Hell's Kitchen / American Data Power
CTR's Hell's Kitchen project targets 1,100 MW of baseload geothermal power from the Salton Sea field, with a Stage 1 final investment decision targeting 50 MW paired with lithium extraction. In September 2025, CTR and Baker Hughes announced definitive agreements to develop up to 500 MW of geothermal power specifically to serve hyperscale data centers. CTR's CEO Rod Colwell stated the intent explicitly: the company hopes data centers "will set up shop in Imperial County alongside their geothermal operations" to have "direct access to the energy they need." In January 2026, CTR launched American Data Power — a dedicated subsidiary — to deliver a 600 MW geothermal energy complex for AI data infrastructure, co-located in Lithium Valley. The project is FAST-41 designated, ensuring interagency coordination to expedite federal permitting. The geothermal resource has a 40-year operating history in the field.
CalEthos / TerraVolt Infrastructure
CalEthos is developing a 420 MW data center in Imperial Valley's Lithium Valley, explicitly planned to run on 100% geothermal and solar energy. In May 2026, CalEthos launched TerraVolt Infrastructure Inc. as a dedicated behind-the-meter geothermal subsidiary. TerraVolt describes itself as an Infrastructure-as-a-Service platform providing "turnkey solutions to hyperscalers, colocation providers, and data center developers seeking an expedited route to market" via behind-the-meter geothermal power. CalEthos's own description: its 420 MW facility will be powered by 100% renewable energy — the behind-the-meter model, not CAISO grid power.
The DOE's Explicit Endorsement
The U.S. Department of Energy's Office of Geothermal has formally identified the geothermal-data center co-location opportunity. Its analysis states that geothermal's 90% capacity factor "allows geothermal power plants to operate 24 hours a day, with steady output nearly all of the time" — making it "ideal for supporting the energy load generated by constant data center operations, such as that required by some AI data computing." The DOE is actively researching Cold Underground Thermal Energy Storage (Cold UTES) as an additional geothermal application for data center cooling, which would further reduce peak electricity demand at co-located facilities.
What Happened to BHE Renewables?
Berkshire Hathaway Energy Renewables suspended its three Salton Sea geothermal projects (Black Rock, Morton Bay, Elmore North — 77 MW each) in February 2025, citing permitting delays and the absence of power purchase agreements. The suspension is relevant but not disqualifying. BHE's projects were designed to sell power to the CAISO grid through utility offtake agreements — the traditional model. They ran into both the transmission interconnection constraint (needing CAISO access) and the PPA market challenge (no utility was contracting for new geothermal at the time). The co-location model being pursued by CTR and CalEthos avoids both problems: the power sells directly to the data center operator, and it never needs to reach the CAISO grid. BHE's projects, tellingly, stalled on exactly the transmission infrastructure constraints that the Powerlink is supposed to solve — but the co-location developers are building around that constraint rather than waiting for it to be resolved.
America's digital economy is at an inflection point. Hyperscale data center and AI demands are surging, but they cannot run on intermittent renewables. The Hell's Kitchen project will provide 500 MW of baseload energy to meet this demand.
How This Rewrites the Powerlink's Demand Case
The two paths — traditional grid connection versus co-located geothermal — produce entirely different infrastructure implications.
Data Center on CAISO Grid
- Connects to SDG&E or SCE transmission system
- Draws power from the CAISO market
- Adds load to Southern California transmission corridors
- Contributes to CAISO congestion and need for new transmission
- Transmission upgrade costs socialized to all CAISO ratepayers
- 4–7 year CAISO interconnection queue
- Powerlink needed? Yes — or adds pressure for it
Data Center Behind the Meter
- Connects to IID's local grid or directly to geothermal plant
- Generates its own power on-site from geothermal
- Never draws from CAISO transmission system
- Adds zero load to SDG&E corridors
- No FERC cost socialization — ratepayers not exposed
- IID interconnection: months, not years
- Powerlink needed? No
The question for the Powerlink's need case is which of these two models dominates the Imperial Valley data center market. If co-location succeeds at the scale CTR and CalEthos are projecting — hundreds of megawatts of committed capacity — then the data-center-driven load growth that partly justified the Powerlink does not materialize on the CAISO system. CAISO's demand forecasts, which are based on utility interconnection applications, cannot easily account for behind-the-meter load that never files for grid service. The CEC's official 4.9 GW data center demand forecast by 2040 reflects only what shows up in the interconnection queue. Co-located, behind-the-meter data centers do not.
What Remains of the Powerlink's Justification
With the data center load removed or substantially reduced, the Powerlink's case rests on a narrower but still real set of arguments. It is important to assess these honestly, because some are strong and some are weaker than SDG&E's presentations suggest.
| Argument | Strength After Co-Location | Analysis |
|---|---|---|
| Grid reliability — Sunrise at capacity, need for redundancy | Strong — unchanged | Sunrise Powerlink is genuinely at full capacity. San Diego's load-pocket vulnerability is real and independent of data center growth. A second 500 kV corridor provides operational resilience during PSPS events. Co-location doesn't address this. |
| Moving utility-scale renewable energy to coastal load centers | Moderate — partially affected | If geothermal is consumed behind the meter by data centers, it doesn't reach San Diego residential customers. The Powerlink would still move solar, wind, and grid-connected geothermal (like the existing 403 MW) to the CAISO system — but the incremental benefit depends on how much new geothermal reaches the grid rather than going behind the meter. |
| Unlocking stranded geothermal for California's grid | Weakened — co-location bypasses this | The prior analysis suggested the Powerlink would unlock stranded geothermal. But CTR and CalEthos are unlocking that same resource without the Powerlink — by routing power to data centers rather than the grid. If the geothermal goes behind the meter, the Powerlink isn't the mechanism that brings it to market. |
| AI data center load growth driving Southern California demand | Significantly weakened | This argument was always partly speculative and partly circular. If the most credible Imperial Valley data centers use co-located geothermal on IID's system, they neither appear in CAISO's load forecast nor require CAISO transmission capacity. The demand growth assumption behind this argument does not materialize on the grid. |
| Electrification demand growth — EVs, heat pumps, industry | Strong — unchanged | This is separate from data centers. California's aggressive electrification mandates for transportation and buildings create real load growth that co-located geothermal does not address. Coastal population centers will draw more power through transmission corridors regardless of what happens in Imperial Valley. |
| Wildfire resilience through geographic redundancy | Strong — unchanged | A geographically distinct third corridor provides PSPS resilience independent of data center considerations. This argument stands on its own merits and is not affected by the co-location model. |
The net assessment: the Powerlink retains a legitimate case based on San Diego's genuine load-pocket vulnerability, electrification demand growth, and wildfire corridor resilience. What it loses with the co-location model is the amplified urgency derived from data center load growth — a factor that SDG&E and CAISO have allowed to color the project's necessity without formally acknowledging it.
The Bubble Risk, Reassessed
The previous analysis warned that an AI bubble burst could leave utility ratepayers holding stranded transmission infrastructure paid for by speculative data center demand. The co-location model modifies this risk profile in a meaningful way — in both directions.
The ratepayer risk is reduced. A data center that builds its own behind-the-meter geothermal plant and then goes bust takes its own stranded assets with it. The geothermal plant was private capital, not rate-base. The IID distribution infrastructure that served it is locally governed and locally financed, not socialized across CAISO. The CAISO ratepayer exposure that the Little Hoover Commission and the Public Advocates Office warned about does not arise for behind-the-meter projects on IID's system.
The Powerlink stranded asset risk, however, is heightened. If the Powerlink is built partly on a demand forecast that assumed data center load growth on the CAISO system — and that load growth instead materializes behind the meter on IID's system — the Powerlink is built for load that was always going to be served by a different infrastructure architecture. It becomes stranded not because AI demand failed to materialize, but because the load materialized in a way that didn't need it. This is a more insidious form of stranded asset risk than simple demand disappointment, because it is not detectable from utility interconnection queue data.
CAISO's demand forecasts are built from utility interconnection applications and CEC load models. Behind-the-meter co-located generation paired with co-located load is largely invisible to this forecasting apparatus. If 500 to 1,000 MW of Imperial Valley data center demand is served by co-located geothermal on IID's system, CAISO's 2022–23 transmission plan — which justified the Powerlink — may have materially overstated the load growth that requires new CAISO transmission capacity. Neither CAISO nor SDG&E has publicly reconciled their demand forecasts with the co-location build-out now underway.
What the Honest Geothermal Picture Means for San Diego
The corrected analysis points toward several conclusions that were absent from the previous version.
For California's clean energy future, the geothermal co-location model in Imperial Valley is genuinely good news. The state has a massive proven resource that has been stranded for decades by the economics of building large geothermal plants to serve distant coastal customers through congested transmission. AI data center demand — by creating a large, local, price-insensitive buyer for firm 24/7 baseload power — may finally make the development economics work. California could end up with hundreds of megawatts of new zero-carbon baseload power that would not otherwise have been built, paid for by tech companies rather than ratepayers.
For the Golden Pacific Powerlink's rationale, the picture is more complicated. The project has legitimate justifications that survive the co-location analysis: San Diego's load-pocket risk is real, electrification demand growth is real, and geographic corridor resilience has genuine value. But the data center urgency that has colored the project's presentation — and that has driven the accelerated timeline pushing toward a CPUC filing before year-end — is materially overstated if co-located geothermal becomes the dominant model for Imperial Valley data centers.
The CPUC proceeding, when it opens, should formally require CAISO and SDG&E to account for behind-the-meter co-located load in their demand projections. If 500 to 1,000 MW of projected Imperial Valley data center demand is already being planned as IID-served, behind-the-meter load, the case for a $1.3 to $2.3 billion 500 kV transmission line is meaningfully different from the case SDG&E presented at its May 2026 open houses.
Where will the power come from? For Imperial Valley data centers specifically, the credible clean answer is geothermal co-location — not natural gas. The Salton Sea field has a 40-year operating history, 1,850 MW of untapped developable capacity, a 90–98% capacity factor, and multiple active developers now explicitly targeting AI data center co-location as their business model. This is not speculative. It is being permitted and financed.
Does it need the Powerlink? No. Co-located geothermal data centers on IID's system require no CAISO grid connection, no SDG&E transmission, no Powerlink corridor, and expose no CAISO ratepayer to cost socialization. The power is generated, consumed, and paid for entirely within a self-contained local system. This is the architecture that eliminates both the 24/7 power problem and the ratepayer equity problem simultaneously.
What does this mean for the Powerlink? The project retains legitimate justification for San Diego's grid reliability, electrification demand growth, and wildfire corridor resilience — but it loses the amplified urgency provided by data center load growth. If California's regulatory process were working correctly, CAISO's demand models would be formally updated to reflect behind-the-meter co-location before the Powerlink's need case is adjudicated by the CPUC. That update has not happened. The CPUC proceeding is the venue where this gap should be formally placed on the record.
For San Diego ratepayers: the geothermal co-location model, if it succeeds at the scale being projected, is the best possible outcome. The data centers get built. The geothermal resource gets developed. California gets clean baseload power. The AI boom is served without natural gas. And the cost lands on tech companies and private investors rather than on residential ratepayers already paying 45.7¢/kWh. The Powerlink, if built, should be sized and justified for what is actually needed — not for a data center demand surge that is being served by an entirely different infrastructure model.
Sources
-
Primary — CTR / Baker Hughes
Controlled Thermal Resources. (September 9, 2025). "CTR and Baker Hughes to collaborate on 500 MW Hell's Kitchen geothermal power project." CThermal.com / ThinkGeoEnergy. [Explicit data center co-location intent; 98%+ capacity factor; 500 MW target.]
https://www.cthermal.com/latest-news/ -
Primary — CalEthos / TerraVolt
Data Center Dynamics. (May 2026). "CalEthos launches geothermal subsidiary, targeting behind-the-meter power for data centers." [TerraVolt IaaS platform; 420 MW data center on 100% geothermal + solar; Rhodium Group 55–64% figure.]
https://www.datacenterdynamics.com/en/news/calethos-launches-geothermal-subsidiary/ -
Analysis — Rhodium Group (via Data Center Frontier)
Data Center Frontier. (2025). "Why Geothermal Energy Could Be a Behind-the-Meter Game Changer for Data Center Power Demand." [Rhodium Group: 55–64% of hyperscale growth could be met behind-the-meter geothermal; 15–17 GW capacity.] datacenterfrontier.com.
https://www.datacenterfrontier.com/energy/article/55274889/ -
Official — U.S. Department of Energy
U.S. Department of Energy, Office of Geothermal. (2026). "Geothermal and Data Centers." [90% capacity factor; Cold UTES cooling; DOE formal endorsement of geothermal-data center co-location.] energy.gov.
https://www.energy.gov/hgeo/geothermal/geothermal-and-data-centers -
Local Analysis — Our Imperial Valley
Our Imperial Valley. (February 5, 2026). "Geothermal Power and AI: Why Imperial Valley Is the Best Location for This Project in America." [IID independence from CAISO; interconnection speed advantage; land cost analysis.] ourimperialvalley.com.
https://www.ourimperialvalley.com/geothermal-power-ai-imperial-valley-best-location-america/ -
News — KPBS
KPBS. (February 12, 2026). "What one company's shift towards data centers says about Imperial County's lithium industry." [CTR CEO Rod Colwell quotes on co-location intent; 600 MW American Data Power complex; lithium industry context.] kpbs.org.
https://www.kpbs.org/news/environment/2026/02/12/ -
Primary — CTR American Data Power
Calexico Chronicle. (January 30, 2026). "CTR launches American Data Power in Lithium Valley." [600 MW energy complex for AI data infrastructure; Hell's Kitchen 4,000 acre project; co-location model.] calexicochronicle.com.
https://calexicochronicle.com/2026/01/30/ -
Wikipedia — Salton Sea Resource
Wikipedia contributors. (2026). "Imperial Valley Geothermal Project." [2,950 MW estimated potential; 2,250 MW currently developable; 403 MW operating; BHE Renewables and EnergySource operators.]
https://en.wikipedia.org/wiki/Imperial_Valley_Geothermal_Project -
BHE Suspension — GPS Business Insider
GPS Business Insider. (July 2025). "BHE Renewables Suspends Three Salton Sea Projects Amid Regulatory Delays, Transmission Hurdles." [Black Rock, Morton Bay, Elmore North; 77 MW each; PPA and transmission interconnection obstacles; IID context.] gpsbusinessinsider.com.
https://gpsbusinessinsider.com/bhe-renewables-suspends-three-salton-sea-projects/ -
Analysis — Data Center Knowledge
Data Center Knowledge. (February 10, 2026). "Geothermal Energy at Scale: A New Paradigm for Data Centers." [EGS technology; co-location site strategy; transmission queue avoidance; Fervo Cape Station 100 MW first commercial EGS.] datacenterknowledge.com.
https://www.datacenterknowledge.com/energy-power-supply/ -
Official — DOE Geothermal
U.S. Department of Energy. (n.d., accessed 2026). "Chapter 2: Geothermal Takes the Stage." [CPUC 2021 procurement order for 1,000 MW firm renewable at 80%+ CF; modern geothermal 90–95% CF; Casa Diablo IV 30 MW; Hell's Kitchen 50 MW development details.] energy.gov.
https://www.energy.gov/eere/articles/chapter-2-geothermal-takes-stage -
Regulatory — BLM California
Bureau of Land Management. (2026). "Geothermal Energy — Regional Information: California." [Salton Sea 2,200 MW estimated capability; The Geysers 1,800 MW; lease management; 10-year lease terms.] blm.gov.
https://www.blm.gov/programs/energy-and-minerals/renewable-energy/geothermal-energy/regional-information/california
Comments
Post a Comment