The case
The application that gets picked.
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.
$150M/year on 50 MW. +$42M on 64 MW. +$54M on 82 MW. Recip 22 MW.
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.
Use the generation permit you already have. Dirty cap 54 MW unchanged. Recip 22 MW.
How the filing stays inside a permit you already have.
One way in
Fastest time to compute
50 MW IT per 100 MWdc
Buy or retask a 2012–2021 plant. The generation interconnect is already permitted. Overlay cooling on 10% of tables — radiator tables, not a new chiller yard. Remaining PV stays generation. Drop in modular AI. Pecos is the worked example; the same screen applies wherever a large remote array is already connected.
Same 100 MWdc fence
+50 MW AI · PUE 1.25 → 1.08
Phase 2 books TES and +14 MWe — 64 MW. Phase 3 extra halls +18 MWe — 82 MW. Recip 22 MW. Dirty cap 54 MW. Same generation permit. Same interconnect. No new water right. No new chiller-yard air permit.
Grid clock — Texas is the named example
A generation interconnect is not a load service. That split exists on every wholesale market. ERCOT West Texas is the named farm example: 50 MW of IT draws 54 MW at the meter at PUE 1.08 — modeled facility load, not a site cap — under a 75 MW large-load threshold. Overnight is a new request. If a 54 MW engine yard will not permit, recip stays 22 MW. Phase 3 still spends +18 MWe on extra halls inside that dirty cap. Campus +47 MW inside 375 MW is the other example.
| If the grid operator… | The site | Grid MW | Time |
|---|---|---|---|
| Approves firm overnight service for 54 MW | Stay grid-tied. Cooling overlay only. No on-site engines. | 54 | Months–~18 months |
| Approves interruptible or flexible service only | Grid by day. Islanded generation covers the night. | 0–54 | Overlay now; engines if required |
| Overnight study is delayed in the large-load queue | Islanded operation. 50 MW of IT still runs. | 0 | Overlay plus generator permit — not a new generation interconnect |
| The alternative is a greenfield hall | Still an overnight study, often a new line. Same 54 MW, no existing interconnect. | 54 | Years if transmission is new |
Texas figures. The same pattern holds on other grids: stay under that market’s large-load threshold where it exists. Heat recovery keeps a backup generator plant smaller — about 22 MW in summer and 28 MW in winter, not a 54 MW engine yard. That residual plant is what air, noise, and water will consider. Growth past 50 MW is recovered heat, not a bigger stack.
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 MW | Fuel exposure | Gen. water | Cooling water | Time | Emissions / noise |
|---|---|---|---|---|---|---|
| Komorebi cooling-first | $0.72–0.78M | Low | None added | Low | 6–12 mo | Low |
| Reciprocating gas engines | $1.5–2.2M | Medium | Medium–high | Mechanical cooling | 12–24 mo | Medium |
| Komorebi full platform | $3.2M | Low | None added | Low | Staged | Low |
| Aeroderivative simple-cycle GT | $3.2–4.6M | High | High | Mechanical cooling | 24–36 mo+ | High |
| Fuel cells | $4.9–8.0M | High | Low | Site cooling still required | 18–30 mo | Low 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.
Example cover letter
Illustrative letter. Names are fictional. The operating facts are the Pecos case: 50 MW of IT at an existing solar interconnection, peak import 54 MW — under ERCOT’s 75 MW Large Load threshold.
Sierra Mesa Computing, LLC
Grid and Energy
2200 Post Oak Boulevard, Suite 1400
Houston, Texas 77056
28 August 2026
Interconnecting Transmission Service ProviderCopy: Qualified Scheduling Entity for Sierra Mesa Computing
Re: Co-located computing load under 75 MW — Mesa Flats Solar, Pecos County, Texas.
Load Information Form and request for TSP coordination.
Dear Sir or Madam:
Sierra Mesa Computing, LLC requests interconnection of 50 MW of computing load co-located with Mesa Flats Solar, an existing Generation Resource in Pecos County. Peak demand at the point of interconnection is approximately 54 MW, including site auxiliary. Enclosed is the Load Information Form for co-located load under 75 MW, for coordination by the interconnecting TSP and ERCOT.
This is a differentiated load addition at an existing generation interconnection. The plant is already in service. The hall uses power the array is already producing. Peak import remains below ERCOT’s 75 MW Large Load threshold. The load will be registered as a Controllable Load Resource.
We request that the interconnecting TSP coordinate this addition as co-located load under 75 MW, identify any studies required before energization, and confirm the path with ERCOT Resource Integration. We will provide the load model applicable to facilities 25 MW or greater, and any water, cooling, and community-impact information the Commission or ERCOT requests of computational load.
We are prepared to brief the TSP and ERCOT on the net-interchange profile, controllable-load capability, and the closed-loop cooling design.
Respectfully submitted,
Alexandra V. Chen
Vice President, Grid and Energy
Sierra Mesa Computing, LLC
Enclosure: Load Information Form — co-located load under 75 MW
cc: Qualified Scheduling Entity
Modeled on ERCOT’s Load Information Form for co-located load under 75 MW. The interconnecting TSP files with ERCOT. Names are fictional.
