Clear · ΔT > 5 °C
- Same·Point hybrid92
- Dry cooler + adiabatic88
- Dry cooler + chiller95
- Radiative panels48
The case
The application that gets picked.
Cooling performance
The interconnect approval queue decides who wins compute capacity — not GPUs. Undifferentiated 50 MW interconnect filings wait. You are not buying solar. You are buying an existing grid connection and filing a load the operator can say yes to. Texas (Pecos) is the worked example. Every grid has the same two hurdles: getting power out, and getting approval to draw power in after dark.
- CFO
- Economics
- CTO
- Cooling & generation
- Counsel
- Permitting
About $175.5M, then $369.6M, then $405.2M a year.
50.7 MWth. A 250 MWh cold storage tank.
Inside the existing fence. Phase 3 is full off-grid.
Phase 1 cooling. The hourly bars are a 50 MW hall. Phase 2 and Phase 3 are not drawn on them.
Pecos, Texas temperature envelope
Monthly air temperature. Left axis °C, right axis °F. Highs ~17 °C / 63 °F in January to ~37 °C / 98 °F in June–July.
- High
- Mean
- Low
Clear vs cloudy hours
Share of time. Clearer hours help dry and radiative duty; cloud still matters in winter. NSRDB 2024.
- Clearer
- Cloudier
More compute. Naturally.
Cooling performance
Pecos, Texas · challenging weather, firm rejection
The cooling plant is fanless cassette rows, 3.1 MWth of buried pipe, and a 250 MWh cold storage tank. On 17 June at 15:00 the tank discharges at its 35 MWth limit. The chiller runs on surplus PV.
Range across the year
The bars are the year, for a 50 MW hall. On 17 June at 15:00 the chiller supplies 2.1 MWth. Nameplate is 8.45 MWe at COP 6.
Same·Point hybrid
3–5.3 MW, median 3.73.7MW
Dry cooler + adiabatic
7.3–11.4 MW, median 8.48.4MW
Dry cooler + chiller
10.2–15.6 MW, median 11.811.8MW
Radiative panels
2.9–5 MW, median 3.53.5MW
Bar is the inner half of months. Whisker is the full monthly range. Number is the median, MW.
Does the sink still work when weather turns?
Firm heat-rejection capability, 0–100. A chiller always rejects — and spends the watts you wanted for compute.
Cloudy · ΔT > 5 °C
- Same·Point hybrid86
- Dry cooler + adiabatic82
- Dry cooler + chiller93
- Radiative panels22
Clear · ΔT ≈ 0
- Same·Point hybrid78
- Dry cooler + adiabatic28
- Dry cooler + chiller90
- Radiative panels62
Cloudy · ΔT ≤ 0
- Same·Point hybrid70
- Dry cooler + adiabatic18
- Dry cooler + chiller88
- Radiative panels8
Tight band
The cold storage tanks cover the hours the rows cannot and discharge up to 35 MWth. The chiller adds 2.1 MWth and draws 8.45 MWe from surplus PV. It does not wait until the tanks are empty.
Cheap on kWe
Radiative demand looks low until clouds remove the sink.
Pecos, Texas problem set
Clouds plus low or negative approach. Hours that decide available cooling — and therefore compute.
Supplemental chiller
The chiller nameplate is 8.45 MWe at COP 6. On 17 June at 15:00 it supplies 2.1 MWth of the 50.7. It runs on surplus PV.
Left: electricity to hold the same cooling duty, in MW. Right: whether rejection remains available. Illustrative Pecos, Texas envelope — not a live plant meter.
Tracker height · vs fixed tilt
PV tracker fills the shoulders. Cooling tracker fills noon, stall, and night.
June 17, Pecos, apparent solar time (noon is 12:00). Same cassette. A SAT PV table buys morning and evening kWh (32% POA at 09:00 vs 25° south, 2% at noon). Inverse-track radiators do the opposite: they turn off the beam when the fixed rack is hottest. Stall-hour shed is 32% above a 25° fixed-tilt table; night zenith still 9% above a rack that never faces the sky. Civil time in June is CDT — solar noon there is ~13:50.
Cooling · kWth per module · June 17 solar time
kWth
- Tracker · inverse / zenith
- Fixed tilt · 25° south
- 12–16 solar
PV analogue · POA W/m² · SAT vs 25° south
W/m²
- SAT tracker
- Fixed 25° south
- 12–16 solar
Band is 12–16 solar (post-noon heat). Axes are apparent solar time — noon is 12:00, not CDT. Air still lags ~2 h, so the fixed rack’s worst hour is ~14:30, not solar noon. The tracker inverse-tilts off the beam. Classic SAT: +32% at 09:00 solar, only +2% at noon. Cooling does not want that extra beam — it inverse-tracks, so the gain shows up where PV gain disappears.
50.7 MWth · 250 MWh tank
The tank covers the hot hour. The rows cover the night.
At 03:00 on 17 June the cassette rows carry 47.6 MWth and the buried pipe carries 3.1. At 15:00 the rows fall to 10.5, the pipe stays at 3.1, and the 250 MWh tank discharges at its 35 MWth limit. The chiller adds 2.1 MWth straight to the hall. Return water is 45 °C.
| 17 June | Rows | Buried pipe | Tank | Chiller | Total |
|---|---|---|---|---|---|
| 03:00 | 47.6 | 3.1 | 0 | 0 | 50.7 MWth |
| 15:00 | 10.5 | 3.1 | 35 | 2.1 | 50.7 MWth |
17 June is the hottest of the three published days. The tank holds 250 MWh and discharges no more than 35 MWth. The chiller is 8.45 MWe at COP 6 and runs on surplus PV.
The chiller runs on surplus PV.
The chiller is 8.45 MWe at COP 6. It runs on surplus PV, including at the hot hour, when it also puts 2.1 MWth straight into the hall. It is not a 13 MWth machine, and it is not off on 17 June.
The module, separate from the plant
One module is 2.384 by 1.303 metres, 18 tubes, film on the tube and the fin only. At 70 °C in and 0.022 litres per second, the night coil is 2.72 kilowatts, plus 0.17 from the buried pipe. On the hottest afternoon the coil is 1.54 kilowatts. The plant uses a higher flow and a 45 °C return. Do not multiply the module case into the 50.7 megawatt chart.
The business point
A full chiller plant is harder to permit and uses more of the power the GPUs need. This design keeps a chiller, much smaller, for the hours the rows cannot finish.
Most of the heat still leaves on the open radiators. The electrical capacity that no longer runs a full chiller plant can stay with the computers. That is an estimate, not contracted IT.
| Hour | Tracker | Fixed tilt 25° | Tracker / fixed |
|---|---|---|---|
| Stall · 18,239 modules on 40 acres | 1.47 kWth · 26.7 MWth | 1.09 kWth · 19.9 MWth | 1.34× |
| June night | 2.78 kWth · 50.7 MWth | 2.53 kWth · 46.1 MWth | 1.10× |
| Wind at the fins (stall) | 3.13 m/s | 2.09 m/s | 1.50× |
| Sky view | 0.79 stall · 0.90 night | 0.75 all hours | zenith vs 25° |
| Air at the cassette | 44.6 °C stall | 47.1 °C stall (+2.5 K skin) | superadiabatic |
| Acres to close 50 MWth stall (plates only) | 75 acres | 101 acres | 1.3× land |
Physics mix · same energy balance
How the heat actually leaves.
Most of the heat leaves on the wind — air moving over the SPACECOOL lattice and the crown of the buried pipe. A smaller share radiates straight to the cold sky from those same surfaces. The rest soaks into the soil through the buried half of the pipe. Sunlight on the fins is extra heat to dump, not cooling, so it is not in the percentages.
- Convective 69% / 58.5 MWth yr-mean
- Radiative 16% / 13.9 MWth
- SPACECOOL ε=0.95
- Conductive 15% / 12.6 MWth
- buried Ø800
MWth · stacked monthly
18,239 radiator modules, 47,530 m Ø800 (radiator rows only), Pecos NSRDB monthly envelope. Shares are year-mean of each mode ÷ (convective + radiative + conductive). Convection is the main cooler. July is the stall trough — do not flatten it. Solar on the fins is omitted from the stack.
Land · same acreage
Same site. Phase 1 cooling rows. Later phases share a larger reject field.
Phase 1 keeps the PV and puts about 35 acres into open cassette rows, plus an air-cooled chiller. No engine yard. Phase 2 and Phase 3 use one shared reject field, about 76 acres then 94 acres. Jacket water sets that size. The engines take their own pad: about 2.3 acres, then 3.2. That is the equipment yard, not the noise setback. The condenser alone would be about 36 acres, then 67. The old 137-acre mirror slice is not this field.
A gas field, or a line built for this plant. That is the supply the engine phases need.
Each bar is the same 500-acre site. Gold is the shared reject field: 35 acres in Phase 1, 76 in Phase 2, 94 in Phase 3. The engine yard is a separate pad: none, then 2.3 acres, then 3.2. A G3520H bay is 16 m by 6 m. A 34SG bay is 18 m by 8 m. The yard is twice the bays, for fuel, switchgear, stacks, and a fire lane. It is not the 55 dBA setback. Jacket water still sets the reject field. The rest of the acreage is not a mirror field.
The ground
An optional sink. Not the whole plant.
The buried pipe is a 3.1 MWth credit on the hottest hour and at night. It does not replace the chiller, and more pipe does not.
24-hour example
Where the power and the cooling come from.
17 June is the hot day. 6 December is the cold day. The top plot is the electricity into the 50 MW hall. The lower plot is the cooling, and the solar heat when the mirrors are running. Phase 1 firm compute is 58 MW. Phase 2 is 153 MW and stays on the grid. Phase 3 is 175 MW and leaves the grid. These hours are the hall. They are not a redraw of the whole plant.
- PV to the hall
- Battery
- Grid
- Cassette rows
- Buried pipe
- Cold storage
- Chiller
- 50 MW hall
Compared with
Heat has to go somewhere.
Four substitutions a facilities team will offer. The stall day is the test — not PUE in isolation.
Warm-water liquid cooling and dry coolers
That path can reach a similar PUE on a campus that already has power. On a stall afternoon outdoor air matches the loop. Dry coolers stop. Adiabatic buys those hours with water the permit often will not give.
More batteries
Batteries store electricity. They do not reject heat. Every MWh spent running chillers is a MWh that never reaches GPUs.
Chillers on midday surplus
Noon PV can cover the chiller. After sunset the array is dark. On 17 June at 15:00 the chiller supplies 2.1 MWth of the 50.7, not the whole load. Nameplate is 8.45 MWe at COP 6.
CDUs and liquid-to-chip
They move heat out of the rack. They still need a place to dump the heat. Phase 1 uses about 35 acres of cassette rows. Later phases share about 76 acres, then 94.
Why the overlay beats a new plant
- Reuses the interconnect and land you already have
- About 35 acres of radiator rows, plus a smaller chiller, reject 50.7 MWth for the 50 MW hall
- The chiller nameplate is 8.45 MWe at COP 6. On the hottest hour it supplies 2.1 MWth of the 50.7.