← All docs

The load and the grid — a city’s worth of demand, and who carries it

A #233 extension narrative. Like end-use-and-workloads.md and defense-nexus.md it is hand-assembled analysis over cited records, and it ends at the questions it cannot close. Every statement carries its register: [verified] (read from a cited record or a connector pull), [inference] (a labeled derivation), [open] (a question the record does not answer), [reference] (an outside-published figure or law). Two disciplines govern this page, because the two phrases people reach for are both softer than they sound. The headline “313 MW” is backup generation, not the operating load — and the number behind it is redacted in the final permit. And “behind-the-meter” is a proponent claim, not a documented fact about this campus, which the record classifies as a grid-served retail customer. What survives both disciplines is still large, and still the point: a single campus is a material fraction of its utility’s entire load, for a headcount a big-box store would exceed.

The plainest way to say what this report is: the campus is a very large electricity customer — on the order of a mid-size city’s demand, on one corridor — and it is served off the grid by AEP Ohio inside the PJM market. That much the record carries. The two things most often said about that load — its exact size, and whether it runs “behind the meter” — are where the record gets thinner than the confidence around them, and where the public-cost question actually lives. This walks the solid part, the soft part, and the part that isn’t in the record at all.

———

the number everyone cites is the backup — and it’s redacted

The figure in every summary is ~313 MW. It is real, and it is worth being precise about what it measures. It comes from the campus’s Ohio EPA air permit: 114 diesel emergency generators × ~2,750 ekW each ≈ 313 MW [verified: OEPA Air PTI P0138965, draft — data/extracted/permits/3987141.epa.yaml]. That is standby capacity — the diesels that run when the grid drops — not the power the campus pulls to do its work.

And the per-engine rating behind it does not survive into the final document. The final permit, issued 2026-05-28, redacts the engine make, model, and size as Confidential Business Information / trade secret (Comments 16, 19; Response 16), confirming only the unit count — 114 data-hall gensets plus one smaller HUBGEN, 115 emission units — and the emission rates. The ~2,750 ekW/engine figure behind the ~313 MW total comes from the draft public notice, not the issued permit [verified: data/extracted/permits/4132514.epa.yaml]. So the most-quoted number about this facility’s power rests on a draft the final record withheld — the same withholding-as-evidence pattern that runs through the rest of the corpus.

The operating load is an inference, and it’s labeled as one. Backup capacity on a hyperscale design tracks IT load at roughly N+1, which puts the IT load near ~250–300 MW (midpoint 275) [inference: N+1 backup ≈ IT], and the total facility draw — IT plus cooling and losses, through the power-usage model — near ~348 MW [inference: IT × PUE]. Those are the numbers the rest of this page uses, carried as inferences, not facts. The honest version of the headline is: a standby bank the record sizes at ~313 MW, around a working load the record lets you estimate at a city’s scale but never states.

———

a city’s worth of demand on one corridor

Take the inferred ~348 MW facility draw at a realistic load factor and the annual energy is ~2,740 GWh/yr [inference: derived]. Set against the grid it sits in, the comparison is the finding:

The headline reads off the first row: a single campus equals roughly 5–6% of its serving utility’s entire retail electricity sales [connector]. By a different yard- stick, its ~2,740 GWh/yr is the consumption of ~260,000 Ohio homes [inference: derived, EIA-cited], ~1.8% of all the electricity sold at retail in the state. This is one customer, on one corridor, sized like a small city.

Whether the grid can physically carry it is not the worry. Over a representative summer, PJM’s in-balancing-authority generation runs ~5,747 MW above its own demand on average, so a ~348 MW load sits comfortably inside that headroom without net imports [connector: EIA-930, screening]. The question a load this size raises is not can the power be generated — it can. It is who pays for moving it.

———

load, not jobs

This is the comparison the public record actually substantiates. Set the verified columns side by side [verified: ECONOMICS.md §2, the county's own production]:

• a ~275 MW IT load (inferred) and 3.1–10 MGD of consumptive cooling water, against
• ~50 permanent jobs and ~$4M of payroll by 2030, on a 15-year / 75% tax abatement over a ~$500M build, with $14.2M of roadwork routed through the public Port Authority.

That is on the order of ~5–6 MW per job [inference] — a community-scale electrical load and a multi-MGD basin draw, for a headcount a single big-box store would exceed. The structural argument the corpus substantiates is exactly that: the public subsidizes load and consumption, not employment [inference] — and does so, per the land record, for a counterparty named only as a Delaware shell. It is the demand-side mirror of the water finding: the burden scales with the megawatts, the public benefit scales with the jobs, and the two are orders of magnitude apart.

———

who carries the grid

A load this size has a price the campus sees and a price the public might. The campus’s own footprint in the PJM market, sized against published price signals, is large: roughly $96M/yr in energy at the AEP-zone wholesale price and roughly $34M/yr in capacity at the 2025/26 auction’s clearing rate — a price that spiked ~9× over the prior year [reference: PJM Data Miner / RPM, screening, verify-flagged]. Those are the campus’s costs, not the public’s. The public-cost question is the one underneath them, and it is [open]: who pays for the interconnection and network upgrades a new community-scale load requires — the campus, or the ratepayers on the same system?

The record sets up that question without answering it. The campus is classified a PUCO-regulated retail customer of AEP Ohio — grid-served, not wholesale [verified: FERC seam, classification]. Its ~2,740 GWh/yr lands as ~1.8% of Ohio retail sales, the basis for a deliberately stylized 0.9–1.8% consumer-price-pressure screen that the corpus flags as a sensitivity, not a forecast [inference, low]. The proponents say the cost-causers pay: Google testified it takes service under PUCO- approved tariffs and runs a 100 MW PJM virtual-power-plant; a competitor witness invoked “BYONG” (Bring Your Own New Generation) to “deliver more than we utilize.” But the same testimony carries no per-site disclosure of who operates that way, or what share [verified: proponent-analysis.md]. Whether this campus’s grid upgrades fall on the campus or the rate base is not in the record.

———

the behind-the-meter question

The phrase that travels with this facility is “behind-the-meter,” and it deserves the same discipline as the megawatts. Behind-the-meter co-location — a large load wired directly to a generator, bypassing the grid it would otherwise pay into — is a live and unsettled federal question: FERC rejected the Susquehanna–Amazon co-location amendment (ER24-2172, 2024-11-01) and opened a technical conference on the broader issue (AD24-11-000) [reference: FERC dockets, verify-flagged]. It is real, and it is exactly the kind of arrangement that would move who-pays from PUCO to FERC.

For this campus, it is [open]. The record classifies the campus as grid-served retail. The 114 generators are emergency backup, explicitly not primary generation [verified: air permit]; whether any primary on-site generation exists, and on what fuel, is unproven in the corpus [open]. The developers’ public FAQ lists “behind-the-meter power” among its efficiency claims [reference: AEDG], but a marketing claim is not a documented interconnection. So the honest statement is the narrow one: behind-the-meter is the policy seam this load sits next to, and a posture its proponents market — not an arrangement the record shows for the Lima campus.

———

where this stops

Strip it to what the record will and won’t carry. It will carry the scale: a campus that is ~5–6% of its utility’s entire retail load, ~260,000 homes of annual consumption, for ~50 jobs — load, not jobs, with confidence. It will carry the grid it sits in: AEP Ohio, PJM, a capacity price that just spiked ninefold. What it will not carry is three things the public-cost verdict actually needs:

• the real operating load — the figures are nameplate and inferred, and the nameplate behind the famous “313 MW” is redacted in the issued permit [open];
• who bears the grid cost — whether the interconnection and upgrades fall on the campus or the rate base is not disclosed [open];
• whether any of it runs behind the meter — claimed and marketed, not documented [open].

A megawatt is easy to print and hard to hide; that is why the load is the solid part of this story. Who pays to carry it, and on whose meter, is the part the record keeps — and until that is disclosed, the honest reading is the one the numbers already force: the public is being asked to host a city’s worth of demand for a small town’s worth of jobs, and to take the cost allocation on faith.

———

sources

• The ~313 MW backup, the IT-load and facility-draw inferences, the load-vs-jobs mismatch, and the consumer-price-pressure screen — ECONOMICS.md §1, §3, §6 (hand-assembled over cited records)
• The redaction of the per-engine rating in the final permit — data/extracted/permits/4132514.epa.yaml (Comments 16, 19; Response 16); the ~2,750 ekW draft figure — data/extracted/permits/3987141.epa.yaml
• The serving utility, the grid shares, the PJM headroom and market footprint, and the FERC jurisdictional seam — docs/GRID.md §#94–#97 (epic #93); the datasets under data/reference/eia/, data/reference/pjm/, data/reference/ferc/
• The proponents’ grid-cost and BYONG claims, and the absence of per-site disclosure — legal/proponent-analysis.md
• The abatement, jobs, capital, and roadwork terms — the county’s PRR production, CRA Res #548-25; DOSSIER.md §6
• The water consequence of the same load — HYDROLOGY.md, toxics-and-the-corridor.md