07 · Technical
Three plants, one site, one set of physics
CSP with 12-hour salt, a 30 MW PV farm, a 20 MW binary geothermal unit, RO water and 20 ha of greenhouses. The hybrid is the point. It is also why the capacity-factor stack cannot be 80%.
Sun
GHI 6.1–6.7 kWh/m²·d
CSP 200 MW
Trough + 12 h salt
PV 30 MW
Bifacial, daytime
Geo 20 MW
ORC, 180°C design
Grid 230 kV
Semera corridor
Heat & brine
Greenhouses + RO
Energy accounting
| Block | MW | Sponsor CF | Sponsor MWh | Independent CF | Independent MWh |
|---|---|---|---|---|---|
| CSP + 12 h TES | 200 | 50% | 876,000 | 48% | 840,960 |
| PV bifacial | 30 | 40% | 105,120 | 26% | 68,328 |
| Geo ORC | 20 | 90% | 157,680 | 85% | 148,920 |
| Sum of CFs | 250 | 52% | 1,138,800 | 48% | 1,058,208 |
| Booked in the pack | 250 | 80% | 1,752,000 | — | do not use |
Independent PV 26% reflects Afar heat, soiling and inverter derate vs a 40% marketing CF. CSP 48% assumes good DNI and dry cooling — still unmeasured.
CSP — the $1.3 billion object
IRENA 2024: global weighted CSP LCOE $0.092/kWh; installed cost still spans $3,571–$9,699/kW (4–8 h TES, 2018–19); 8h+ TES $4,574–$7,774/kW. Xina Solar One (South Africa, 100 MW, 5.5 h) cost $880M in 2017 — about $8,950/kW in 2020 dollars. A 200 MW / 12 h Ethiopian first plant at $6.5/W is ambitious, not insane. At $2.50/W it would be fiction; the original pack at $6.00/W is the one line that is roughly in the market.
PV and geothermal
30 MW of N-type TOPCon bifacial is the easy block. Utility PV in 2026 clears well below the pack’s $1.50/W; we use $0.90/W installed in Afar with tracking. Geothermal is the opposite: $2.50/W is KenGen-on-a-known- field economics. A 20 MW Tendaho first plant, including 6–10 deep wells, a gathering system and an Ormat-class ORC, belongs nearer $5–6/W plus a dry-hole contingency. Inlet 180°C / 12.5% net is a plausible binary design only if the produced enthalpy holds.
Independent CAPEX build-up
Total ~$1.88B, coincidentally close to the sponsor $1.86B. The problem is not the headline; it is that geothermal and grid are under-cooked inside it, and 15%+ contingency for a first Ethiopian CSP is a floor.
Comparables
| Project | Tech | Cost signal | Lesson |
|---|---|---|---|
| Xina Solar One (ZA) | 100 MW CSP trough, 5.5 h TES | $880M (2017) ≈ $8,950/kW (2020$) | African CSP benchmark. First-of-kind in Ethiopia should not undercut this on $/kW. |
| Noor Ouarzazate complex (MA) | CSP + PV, multi-phase | Multi-billion, concessional + IFI | Required sovereign and climate-fund scaffolding. |
| Cerro Dominador (CL) | 110 MW tower, 17.5 h TES + PV | High-$/kW tower with long storage | Atacama DNI is world-class; Afar DNI is unproven at that level. |
| Metehara Solar (ET) | 100 MW PV | Sponsor cites ~$1.20/W | Ethiopia solar is real; CSP is not. Do not analogize PV IRRs to 200 MW CSP. |
| Kenya Olkaria (KE) | Geothermal, ~800 MW class field | Typically $4–7M/kW including wells | Closest geothermal cost analogue. Sponsor $2.50/W for Tendaho is below observed African geothermal. |
Farming block
20 ha of geothermal-heated greenhouse can produce high-value vegetables for Djibouti/Semera, not 5,000 t of commodity plus 40,000 t of organic fertilizer. Feedstock for fertilizer at that scale is a municipal or agro-industrial waste contract that does not exist in Dubti. Phase 1 should be 5 ha and a soils lab.