Skip to content
KomorebiEnergy + AI

The case

The application that gets picked.

Buy the interconnect. Permit the cooling. Host the IT.

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. The same two clocks exist on every grid: power out, and permission to pull power in after dark.

CFO
Economics

$150M/year on 50 MW. +$42M on 64 MW. +$54M on 82 MW. Recip 22 MW.

CTO
Cooling & generation

50 MW holds on the hottest design day. 54 MW is the modeled meter draw at PUE 1.08, not a site cap — versus 62.5 MW at PUE 1.25.

Counsel
Permitting

Use the generation permit you already have. Dirty cap 54 MW unchanged. Recip 22 MW.

Firm offtake, flex hours, and what the overlay costs.

The purchase

Buy the interconnect. Overlay cooling. Host 50 MW of firm IT.

This is a real-estate and cooling purchase, not a generation bet. Hang-on overlay $6.4–7.5M. $36–39M to close 50 MW ($0.72–0.78M/MW). Do not call $36M an overlay.

Acquire
100 MWdc, commercial operation 2021 or earlier. The interconnect comes with the deed.
Overlay
$6.4–7.5M cooling overlay on the tables. Dry coolers stay as fail-safe.
Host
50 MW of 99.9% firm AI in 6–12 months. Dry coolers stay as fail-safe.
Earn
Phase 1: $150M/year on 50 MW. Phase 2: +$42M on 64 MW. Phase 3: +$54M on 82 MW extra halls. Recip 22 MW. Dirty cap 54 MW.
Time
A new interconnect is 5–7 years. This plant already exports. Overnight import is 54 MW, not 62.5.

CFO

  • Compare $0.72–0.78M per MW to close the hall against on-site engines and a new interconnect. Overlay itself is $6.4–7.5M.
  • Phase 1 floor is 50 MW — $150M per year at $3M per MW-year. Phase 2 is 64 MW — $192M. Phase 3 extra halls is 82 MW — $246M. Same planning rate.
  • 111 MW / 154 MW / $333M is the interconnect-off stack. It is not this farm walk.
  • Phase 3 is extra halls: +18 MWe, recip still 22 MW, dirty cap 54 MW unchanged. It does not “hold 50 MW.”
  • The $36–39M invoice closes 50 MW of firm AI. The overlay is $6.4–7.5M. It is not a kWh-savings project.
  • Construction follows a 50-acre radiator reservation on the fenceline you already own. Remaining PV stays generation.
  • The farm bid is a separate energy-asset purchase. The overlay is the cooling invoice. GPUs stay on the customer’s books.

CTO

  • Midday surplus can run conventional chillers. After sunset there is no array to run them, and on the hottest afternoons dry coolers cannot reject 50 MWth.
  • When outdoor air is as hot as the loop, 50 acres of radiator tables dump 50 MWth. 50 MW of IT draws 54 MW at the meter at PUE 1.08 — modeled facility load, not a site cap. That draw sits under ERCOT’s 75 MW large-load threshold.
  • At the meter: 54 MW with Komorebi (PUE 1.08) versus 62.5 MW with chillers. Same 50 MW of IT. Dry coolers remain as fail-safe.
  • Phase 1 is grid-tied and thermally firm on a Pecos design day. Phase 2 books TES and 14 MWe on the same loop. Phase 3 spends extra clean MWe on extra halls. Recip 22 MW. Dirty cap 54 MW.

MWdc is the farm you buy and pipe. MW AC is what the interconnect carries. MW IT is the compute you contract.

First commercial plant

Scale from a 10 MW Phase 1 plant.

That is the first plant we contract. Other sizes stay linear with the 100 MWdc existing-PV conversion.

Fast-track AI compute
10 MW
Additional modeled AI compute
+2.8 MW
99.9% firm at maturity
12.8 MW
Peak / flexible
16.4 MW
Repowered PV
20 MWdc
Recovered CSP
18.8 MWth

Illustrative enabled-compute value, at $3M per incremental MW-year: $30.0M / year

Qualify this size

Permitting and economics per additional MW of AI

Capex, water, noise, air, and time are the same decision. Cooling-first reuses a permitted interconnect and avoids a new chiller yard. Time is to additional IT, not to a greenfield campus. Texas is the worked example; the comparison holds on other grids.

Path$M / AI MWFuel exposureGen. waterCooling waterTimeEmissions / noise
Komorebi cooling-first$0.72–0.78MLowNone addedLow6–12 moLow
Reciprocating gas engines$1.5–2.2MMediumMedium–highMechanical cooling12–24 moMedium
Komorebi full platform$3.2MLowNone addedLowStagedLow
Aeroderivative simple-cycle GT$3.2–4.6MHighHighMechanical cooling24–36 mo+High
Fuel cells$4.9–8.0MHighLowSite cooling still required18–30 moLow stack, fuel upstream

Reciprocating engines are typically more efficient than simple-cycle aeroderivative turbines. Combined-cycle GTs are not in this table. Fuel-cell generation-side water is Low in normal operation.

Indicative cost of energy that reaches the halls — $ / MWh

Cooling-first cuts overhead at the meter, so more of each purchased or generated MWh reaches IT. That is not a generation LCOE. Generator rows below are conventional generation LCOE.

Path$ / MWhBasis
Komorebi cooling-first$40–60Delivered to AI compute — overhead reduction
Reciprocating gas engines$55–85Generation LCOE
Komorebi full platform$35–55Delivered to IT — duration and ORC hours
Aeroderivative simple-cycle GT$70–110Generation LCOE
Fuel cells$95–145Generation LCOE

Figures are from an illustrative U.S. planning case: 100 MWdc existing PV farm, 500-acre 2015 fence. Phase 1 books 50 MW of firm IT. Inside that hall, parasitics fall 8.5 MW (12.5 MW of compressors to 4 MW of pumps); meter 54 MW at PUE 1.08, not 62.5. Phase 2 books TES and +14 MWe (64 MW). Phase 3 spends +18 MWe on extra halls (82 MW). Recip 22 MW. Dirty cap 54 MW. Actual results vary by site, resource, fuel, interconnection, and execution. Patents pending. Not an offer.

Worked example · 100 MWdc existing PV · Pecos, Texas

Book 50 MW. Then 64. Then 82. Recip stays 22.

Phase 1 is the underwritten farm case — grid-tied, cooler on 8–10% of tables, 50 MW of design-target 99.9% firm AI at the existing meter. Remaining PV stays generation on GCCM. Leftover acres take Phase 2 CSP on concrete rail. TES is booked in Phase 2: +14 MWe, 64 MW, +$42M/year, same loop, no dry-cooler condenser. Phase 3 spends +18 MWe on extra halls — 82 MW — with recip still 22 MW inside a 54 MW dirty cap. 111 MW is a different stack.

Phase 1 — cooling
+50 MW AI

Design-target 99.9% firm AI. Cooling on 10% of tables. Remaining PV stays generation. First commercial plant.

Phase 2 — TES booked
64 MW AI

TES booked here. +14 MWe on the same radiator loop. 64 MW. No dry-cooler condenser.

Phase 3 — extra halls
82 MW AI

+18 MWe extra halls. 82 MW. Recip 22 MW. Dirty cap 54 MW. Not 111/154.

Starting plant is 100 MWdc on the same land. Re-module holds 100 MWdc of PV on a denser foundation. 40–50 acres convert to radiator tables. Leftover acres take Phase 2 CSP steel. TES is booked in Phase 2. Phase 3 is extra halls.

$150M / year — Phase 1 firm

50 MW of design-target 99.9% firm AI × $3M per MW-year planning value. A planning year, not a live meter.

$192M / year — 64 MW

Phase 2: 50 + 14 MWe. Same $3M rate. TES booked. Same radiator loop.

82 MW — extra halls

Phase 3: 50 + 14 + 18. $246M/year at $3M per MW-year. Recip 22 MW. Dirty cap 54 MW.

Phase 3 is extra halls, not a residual plant that holds 50 MW. 430 °C exhaust goes direct to HT TES. +18 MWe, 82 MW of IT, +$54M/year. Recip stays 22 MW. Dirty cap 54 MW unchanged. 111 MW firm / 154 MW flex / $333M is the interconnect-off stack — a different walk. Not in the Phase 1 offtake.

Eight-day Pecos screen on the radiator campus: 50 MW firm served in the window, PV held at 100 MWdc, 50 acres reserved for tables. A full-year radiator-campus run is not yet the published design target. Prior whole-field radial year is not this plant. A 10 MWdc pilot — about 5 MW of firm IT — is the step that validates hot-day performance, parasitic demand, and ground-temperature behavior at commercial scale.

Compared with

The metric is firm IT.

Two comparisons. First: conversion of an already-connected solar deed into a 50 MW IT host this year — no interconnect and no 24/7 heat sink is 100% of nothing. Second: 8.5 MW less parasitic load versus PUE 1.25 at the same 50 MW IT. Book Phase 1. Treat later MW as options.

  • A new interconnect or a new plant

    A site with no interconnect and no 24/7 heat sink this year is 100% of nothing in IT offtake. That is a 5–7 year clock. This farm already exports. Overnight import is a separate filing. 54 MW is the modeled meter draw at PUE 1.08, not a new generation tie.

  • 8.5 MW less parasitic load

    Same 50 MW of IT: 54 MW at the meter at PUE 1.08 versus 62.5 MW at PUE 1.25. That incremental comparison is second. Releasing 8.5 MW of parasitics is what makes 50 MW contractable at a planning value of $3M per MW-year.

  • On-site engines for the 50 MW

    TES is booked in Phase 2 so islanded shaft is ~22 MW. Dirty cap 54 MW unchanged. Phase 3 spends +18 MWe on extra halls — 82 MW — with recip still 22 MW. Engines last. Not the 111 MW interconnect-off stack.

  • More batteries as the cooling plant

    A pack can run compressors. It cannot dump server heat. Every MWh spent on chillers is a MWh that never reaches GPUs.

Why the overlay beats a new plant

  • Lower capex per additional MW of AI than recips or simple-cycle GTs — and no new water right or chiller-yard air permit
  • Reuses the interconnect and land you already have, or makes a campus you already permit hold more IT
  • The array stays PV. A 50-acre radiator reservation is the 24/7 cooling plant — then AI compute on the interconnect you already have
  • If overnight service is delayed — or the air permit cannot take 54 MW of engines — Phase 2 TES/ORC is already in, so the islanded plant is ~22–28 MW, not a 54 MW engine yard
Request the briefing pack