5502 Spinks Rd · Abilene, TX · settlement LZ_WEST
Crusoe Abilene (Lancium Clean Campus)
Public remote observation established measurable roof reflectance and thermal behaviour and validated the observation → model chain. The whole-system model says the larger recoverable opportunity sits in the cooling plant — if the inferred degradation states exist. No mechanical fault has been observed.
- 1Data center · 11 roofs · 978,237 sf · ERCOT AEP Texas (transmission) / Lancium private
What is this facility?
Observed5502 Spinks Rd · settlement LZ_WEST · 3 cooling-architecture priors
- 221,462 calibrated Sentinel-2 observations and 11,160 Landsat thermal observations, 60,336 modeled system hours.
What does Pulsix currently know?
Observed102,648 weather hours · 100,572 price intervals · latest imagery 2026-09-18 · baseline dc-baseline v1.2.0
- 32 roofs persistently below peers in reflectance; roofs that read darker run hotter on the same day.
What is unusual?
Peer derivedGeneral Roof 4 · DATA_HALL_04 sits 0.069 albedo below its peers, persistently General Roof 2 · DATA_HALL_03 sits 0.047 albedo below its peers, persistently
- 4If the inferred plant-degradation states exist, the modeled recoverable opportunity is MW-scale — coil clean alone ≈ 6.1 MW facility peak
What might it be worth?
InferredEnergy lens $900k/yr for the top intervention; capacity and compute lenses are zero unless the facility is power-constrained (0 % of hours as modeled). Roof cleaning is small.
- 5Roof behaviour is observed (6 observed sources); the plant is modeled on 13 manufacturer/standard and inferred priors; 7 owner-data gaps bound the rest.
How confident are we?
Modeledbands are coherent favorable/unfavorable worlds, not independent errors · Landsat roofs are < 1 native pixel · sun-geometry seasonality not removed from albedo trends
- 6cooling topology, equipment schedule / nameplates; then monthly utility bills or bms it-kw and interconnect limit.
What should we validate next?
Owner data requiredA/B/C configuration prior (±2× on chiller kW, water, PUE)
What public observation established
ObservedMeasured from orbit and cross-checked across sensors. These are observations; none of them names a cause.
- ObservedAvailableSame-day Sentinel-2 and Landsat agree: roofs that read darker than their peers run hotter (r = -0.49, -12.7 °C per unit albedo)
420 matched pairs over 62 acquisitions (landsat-8, landsat-9, sentinel-2a, sentinel-2b, sentinel-2c), ≤1 day apart, both quantities peer-normalized at the same acquisition. Holds in low and high irradiance and both seasons. Every roof is a mixed 100 m thermal pixel — agreement is necessary for a physical reading, not attribution of cause.
- Peer derivedAvailableGeneral Roof 4 · DATA_HALL_04 sits 0.069 albedo below its peers, persistently
69 of 122 clean acquisitions ≥0.03 below the completed-membrane peer median (mean z -2.47), 2024-11-14 → 2026-06-07, on sentinel-2a, sentinel-2b, sentinel-2c. Persistence is a diagnostic target; soiling, weathering, ponding, a different product and mixed pixels are all still open.
- Peer derivedAvailableGeneral Roof 2 · DATA_HALL_03 sits 0.047 albedo below its peers, persistently
75 of 135 clean acquisitions ≥0.03 below the completed-membrane peer median (mean z -1.94), 2024-11-22 → 2026-06-07, on sentinel-2a, sentinel-2b, sentinel-2c. Persistence is a diagnostic target; soiling, weathering, ponding, a different product and mixed pixels are all still open.
- ObservedAvailableGeneral Roof 8 · DATA_HALL_01 is a visually clean-control candidate, not a verified control — it reads below its own age-adjusted expectation
What the whole-system model says
ModeledManufacturer and engineering-standard physics on assumed IT load and inferred as-found states. Read every line as “if this state exists, the modeled recoverable opportunity is X” — no mechanical fault has been observed.
- InferredModeledIf the inferred plant-degradation states exist, the modeled recoverable opportunity is MW-scale — coil clean alone ≈ 6.1 MW facility peak
Whole-system model under the B ac chiller dry cooling prior: 28,947 MWh/yr, $900,423/yr at hourly LZ_WEST prices, $55 per recovered kW. No mechanical fault has been observed at this campus: fouling, loaded filters and scaled fill are INFERRED priors from engineering literature. BAS trends (approach temperatures, filter ΔP) would turn this from modeled to observed.
- ModeledModeledModeled degradations compound: system interaction loss of 404 MWh/yr beyond the sum of individual penalties