Under constructionGoprofile from priorscurated
Stargate Abilene
OpenAI / Oracle / Crusoe · Abilene, Texas, United States · 1,200 MW · online 2026
Flagship Stargate campus; first buildings energized in 2025 with expansion continuing through 2026.
Supply as reported: Grid interconnect with on-site gas turbines; wind in the ERCOT mix
Build cost
$40.6B
Months to live
49
gated by 8 construction phases
EBITDA / yr
$11.6B
Payback
3.5 yrs
IRR
12.2%
CO₂ / yr
2,472 kt
301 g/kWh
Jobs
2,000 / 880
build / permanent
Composite
62
E 69 · Env 46 · C 73
Goon “Baseline as reported” (55% on-site · 27% wind · 18% grid) · Oct 30, 2026 · meta/muse-wonder-1.0
The verdict is GO because Stargate Abilene clears every floor: 12% IRR against an 8% minimum, payback in about three and a half years against a 7-year maximum, a community score of 73 against a 40 floor, and grid load of 23% against a 60% cap. The environmental case is the constraint, scoring 46 against a 45 floor, with emissions of about 2 Mt a year, so the proposed mix must not drift toward the non-viable higher-carbon alternatives. At about $41B of capex for 960 MW of IT load, 672,000 B200 GPUs and about $4B of ten-year NPV, the economics work if utilization and power pricing hold. Land and water are gating items, not footnotes: only 9 ha of open land was measured against 295 ha needed, and cooling depends on municipal or well supply.
Conditions
- Assemble land beyond the 9 ha measured open on site to the full 295 ha required for the facility and its supply before full campus close.
- Secure contracted municipal or well water for cooling, with no surface water measured within sight of the site, and keep water use and discharge within agreed municipal limits.
- Hold the environment score at or above the 45 floor — it is estimated at 46 — by keeping the proposed gas, wind and grid mix and controlling carbon intensity and water demand.
- Keep economics above floors, with IRR at or above 8% and payback within 7 years, and keep grid draw within 60% of regional headroom — it is estimated at 23%.
- Confirm power supply, rack and GPU delivery for each 142 MW phase before energizing the next phase.
Risks
- Environment is only one point above its floor, so any shift to more gas or grid power, higher carbon intensity or weaker water controls could flip the assessment.
- Land assembly from 9 ha measured open to 295 ha required could delay phases, raise cost or shrink the campus.
- Cooling depends on municipal or well water with no nearby surface water measured, creating supply, permitting and community risk in a medium water-stress setting.
- A 49-month, eight-phase build with a peak crew of 2,500 and a GPU refresh in year five exposes the case to labor, supply-chain and technology-price decay risk.
- Emissions of about 2 Mt a year and a 1,133 MW facility load will draw disclosure, grid and local-impact scrutiny even though grid load is within its cap.
Scores
69economic
46environment
73community
For the executive
This is a GO for the 960 MW Stargate Abilene campus, delivering an estimated 12% IRR and payback in about three and a half years at 65% utilization. Protect the thin environmental margin by locking the supply mix, land assembly and water contracts before scaling beyond the early phases.For the policy maker
The campus would bring about 2,000 construction jobs, 880 permanent jobs, about $182M a year in local spend and about $531M a year in tax revenue, with a community score of 73 against a 40 floor. Approval should be conditioned on assembled land, a sustainable municipal or well water supply, published noise and traffic controls, and monitoring of its roughly 2 Mt a year of carbon emissions.For the investor
The proposal clears all floors, with about $41B of capex, about $4B of ten-year NPV, about $12B of annual revenue and a 12% IRR. The principal underwriting risks are the year-five GPU refresh, GPU-hour price decay, and delivery of land, water and firm power on the 49-month path to live.How to build it so the verdict holds
Build the 960 MW IT campus, about 1,133 MW at the facility meter, in eight phases using the as-proposed viable mix of on-site gas, wind and grid, with B200 GPUs at 80 kW rack density. Sequence land, water and firm power ahead of each hall so the GO economics and the narrow environmental margin both hold.
Configuration
| Supply mix | 55% on-site · 27% wind · 18% grid |
| IT load | 960 MW |
| Hardware | B200 · 80 kW racks |
| Crew · automation | 2,500 · 20% |
| Plan scores | 62 · E 69 · Env 46 · C 73 |
| Live in · payback | 49 mo · 3.5 yrs |
| Clears the floors | yes |
Phasing
- Permitting, land assembly to 295 ha, water agreement and community agreement before land close.
- On-site gas supply and wind procurement in parallel with civil works, gated by firm power contracts for the first phase.
- Phase 1: first 142 MW facility hall energized, gated by supply landing, rack delivery and commissioning sign-off.
- Phases 2 to 4: successive 142 MW halls staggered about four months apart, each gated by power availability, GPU delivery and prior-phase utilization.
- Phases 5 to 8: remaining 142 MW halls to reach 960 MW of IT load, each gated by grid headroom, water capacity and environmental compliance staying above floor.
Community commitments
- Local-first hiring for construction trades, backed by the estimated 2,000 construction jobs and 880 permanent jobs, with regular reporting to the municipality.
- Heat reuse or water offset agreement with the municipality, addressing cooling demand of about 12 million cubic metres a year and the absence of nearby surface water.
- Published noise and traffic limits with a named community liaison, complaint response times and periodic public reporting against those limits.
- Commit the estimated $182M a year in local spend through local procurement targets and publish progress annually to sustain the community score of 73, well above the 40 floor.
Energy mixes the advisor searched
| Mix | Viable | Score | Payback | IRR | CO₂ kt/yr |
|---|---|---|---|---|---|
| Renewables + SMR 50% small · 30% solar · 20% wind | yes | 67 | 4.1 yrs | 8% | 120 |
| Hydro PPA + grid 60% hydro · 40% grid | yes | 65 | 3.4 yrs | 13% | 1,355 |
| SMR long-term 70% small · 30% grid | yes | 65 | 3.7 yrs | 11% | 1,018 |
| As proposed (55% on-site · 27% wind · 18% grid) 55% on-site · 27% wind · 18% grid | yes | 62 | 3.5 yrs | 12% | 2,472 |
| Gas + solar 50% on-site · 50% solar | yes | 62 | 3.9 yrs | 9% | 1,770 |
| Solar + wind + grid backup 40% solar · 30% grid · 30% wind | yes | 62 | 4.0 yrs | 8% | 1,004 |
| Gas bridge 70% on-site · 30% grid | no | 60 | 3.3 yrs | 14% | 3,413 |
| Grid + wind 60% grid · 40% wind | no | 59 | 3.6 yrs | 11% | 1,895 |
| Grid only 100% grid | no | 57 | 3.3 yrs | 14% | 3,189 |
| Grid + solar 60% grid · 40% solar | no | 57 | 3.7 yrs | 10% | 1,917 |
Energy what-if
Energy supply mix
Sliders are weights; shares are normalized to 100%.
Live estimate
vs. baselineComposite
62
Economic
69
Environment
46
Community
73
| Metric | This mix | Baseline |
|---|---|---|
| Build cost | $40.6B | $40.6B |
| Months to live | 49 | 49 |
| Blended power | $0.040/kWh | $0.040/kWh |
| Utilization | 65% | 65% |
| EBITDA / yr | $11.6B | $11.6B |
| Payback | 3.5 yrs | 3.5 yrs |
| IRR | 12.2% | 12.2% |
| CO₂ / yr | 2,472,283 t | 2,472,283 t |
| Carbon intensity | 301 g/kWh | 301 g/kWh |
| Water / yr | 11,778,183 m³ | 11,778,183 m³ |
| Grid draw vs headroom | 23% | 23% |
| Land | 295 ha | 295 ha |
| Permanent jobs | 880 | 880 |
| Nuisance index | 48 | 48 |
- Measured open land is 9 ha; the facility and its supply need 295 ha, so land must be assembled beyond the site (sam-3.1).
- No surface water measured within 0.7 km of the site; cooling water depends on municipal or well supply (sam-3.1).
Region profile country priors
| Industrial power | $0.048/kWh |
| Grid carbon | 380 g/kWh |
| Interconnect lead | 14 months |
| Grid headroom | 900 MW |
| Grid reliability | 96% |
| Construction trades | 9,000 @ $33/h |
| Land index | 1.00× |
| Sentiment | 64 / 100 |
| Climate · water | temperate · medium stress |
| Solar · wind | 0.90 · 0.95 |
| Geothermal · flare gas | no · yes |
| People within 5 km | unknown |
Baseline build cost
Site measured by sam-3.1 over a 1.4 km view: 9 ha open land, 1 ha built, 0 ha water, 0 ha solar · Built as 8 phases of 142 MW, staggered 4 months · PUE 1.18 from 80 kW · direct liquid in a temperate climate · GPU-hour rate $3.20 decaying 7%/yr per generation; utilization 65% from supply firmness and fill · Blended power $0.040/kWh, 301 gCO2/kWh · Ten-year cash flow at 10% discount with a GPU refresh in year 5
Scenarios
| Name | Mix | Score | Payback | Verdict | |
|---|---|---|---|---|---|
| Baseline as reportedbaseline | 55% on-site · 27% wind · 18% grid | 62 | 3.5 yrs | Go |
Loading map…