Power & ratepayers
Whose grid this load lands on, and what it does to the households on the same wires.
Every megawatt figure about a data center sits over a denominator, and that denominator is somebody's grid. A real impact study names the grid and cites the source. It then states what the campus load does to the households on the same wires.
Whose grid this is
| Link | Identification | Evidence |
|---|---|---|
| Serving retail utility | AEP Ohio (Ohio Power Company) | [reference] |
| Holding company | American Electric Power (AEP) | [reference] |
| Balancing authority | PJM Interconnection | [reference] |
| Wholesale market (RTO/ISO) | PJM Interconnection (RTO/ISO) | [reference] |
| Retail rate regulator | Public Utilities Commission of Ohio (PUCO) | [reference] |
What those systems already carry
| System | Annual load | Campus share |
|---|---|---|
| AEP Ohio (Ohio Power Company) | 48,653 GWh/yr | 3.16% |
| PJM Interconnection | 815,056 GWh/yr | 0.19% |
| State retail sales | 161,934 GWh/yr | 0.95% |
The disclosed campus draws ~195 MW — at a 0.9 load factor, ~1,537 GWh/yr, the
numerator of every share above. The draw is itself an [inference].
The campus against consumer prices
Applying the stated short-run transmission coefficient ( 0.75 %price/%demand) to that share gives 0.47–0.95% on the 16.96¢/kWh residential price — deliberately stylized, an order of magnitude and never anyone's bill: the campus buys at wholesale, not the residential rate.
- The demand share and households-equivalent are the robust, EIA-cited headline; the price-pressure band is a STYLIZED screening sensitivity, not a forecast of retail bills.
- The campus buys at wholesale/industrial rates, not the residential price shown — residential price is the consumer-impact reference, not the campus bill.
- Facility draw is itself an assumption-laden estimate (air-permit IT load x PUE band, #87); the load factor and transmission coefficient are assumptions.
- Grid foundation layer (#94). The state, AEP Ohio (Ohio Power Company), and PJM denominators are now all connector-sourced — Ohio retail from EIA (shared with #91), per-utility retail from the EIA-861 file, and PJM annual demand from EIA-930. The campus is a single load equal to a material fraction of its serving utility's entire retail sales. Denominator vintages differ (utility EIA-861 2024 / BA EIA-930 2024 / state EIA 2025 (1-year spread)): the EIA state seriesid route publishes ahead of the EIA-861 bulk file and the EIA-930 annual sum, so the three shares are not struck against one common year.
The cited service chain with its full citations, the cohort-price qualification, and the load report's inference chain are in the grid annex.
The disclosed 150 megawatts comes with one more sentence from the same filing: “In 2024, we expanded
the Findlay Megawatt Hub to make available an additional 300 MW of capacity with a standalone
interconnection site.” [verified] Four hundred and fifty megawatts, in a county of 75,000 people,
from a company that withdrew the filing four months later.
Start with the rule, not the reporting
The governing instrument is AEP Ohio’s Schedule DCT, the data-center tariff, at P.U.C.O. No. 22, Original Sheet Nos. 223-1 through 223-7. A great deal written about it is close to right. Four things in the sheets bear directly on this site.
It names cryptocurrency mining outright. The “Mobile Data Center” definition reaches electronic
information services “(including mining of cryptocurrency)” with portable or distributable load
above 25,000 kW. MARA’s contracted 150,000 kW is six times that threshold. [verified]
A Mobile Data Center customer must swear an oath — “a sworn statement, under penalty of perjury,
that neither the customer nor any of its affiliates is owned or controlled by … any foreign
adversary as defined in 15 C.F.R. § 7.4.” [verified] It is the only attestation anywhere in the
schedule.
The +300 MW cannot ride a grandfather. The schedule defines Existing Load that signs a new
service agreement to expand by more than 25,000 kW as New Load; 300 MW is twelve times that
trigger. [inference] — the conclusion follows from the sheets, but the premise, that an executed
pre-tariff agreement exists at all, is [open].
And the famous number is not what it is usually quoted as. “Data centers must pay for 85% of
contracted capacity for 12 years” is too imprecise to quote. The 85% is a floor, applied during a
ramp (Year 1 50%, Year 2 65%, Year 3 80%, Year 4 90%) and afterwards to a rolling 11-month peak or a
Minimum Demand table. Existing Load bills at 60%, not 85%. Term is the ramp plus eight.
[verified]
Read the tariff sheets, not the press summary. Every one of those four points circulates in secondary coverage in a form that is subtly wrong — the threshold, the oath, the percentage, the term. The sheets are public, they are seven pages, and they settle it.
Do not write that the tariff was upheld. Schedule DCT is on appeal to the Supreme Court of
Ohio — In re Application of Ohio Power Co., case 2025-1458, with the Ohio Consumers’
Counsel’s merit brief filed 2026-03-24. It is undecided. The sheets quoted here are the version
filed pursuant to the April 2026 order in AEP Ohio’s distribution rate case 25-392-EL-AIR, not the
July 2025 compliance sheet, which is not captured; any comparison between the two is [open].
Who the customer is, though
Here the tariff runs out. One Power owns the site and its own transmission-voltage substation; MARA
leases capacity behind it. Whether Schedule DCT reaches the host, the tenant, or neither
turns on an electric service agreement nobody has produced. [open] That single unproduced document
would settle three questions at once — who the DCT customer is, whether the site is Existing or New
Load, and what it actually pays.
The record behind this chapter
What this chapter stands on: the records it reads, the inputs its modeled figures rest
on, and the reference data behind its baselines — the same pages the record screens
serve, not a second copy. A figure the record does not support stays [open] and links nothing.
- Grid & consumer energy (US EIA)Reference dataset