A ~340-acre hardened data-center campus is being built in American Township, just outside Lima, Ohio — by a Delaware shell, “Bistrozzi LLC,” for an end user the developers worked to keep quiet. This story shows what the public record reveals, and teaches you how to read it yourself.
Every figure here is checkable.
Each number is read from a cited source and carries a confidence tag, so you never have to take our word for it:
The whole story turns on small numbers
The Ottawa — the river that receives this project downstream, through the small tributaries its outfalls drain to (Pike Run, Dug Run) and the city’s sewers — has a design low flow of about 0.2 cubic feet per second, and runs dry in the driest summers. The campus’s cooling, meanwhile, is drawn not from that river but from off-stream reservoirs a basin this small has to refill. Against that, a campus whose backup power and cooling are built for a city. Holding the document-by-document record against numbers like this is how the analysis moves from anecdote to evidence.
The story draws on 3,382 primary documents, 160 structured records, 89 resolved entities, and 232 dated events — the same feeds the reference library reads.
Each record is read from the image, not the OCR text layer (whose digits are unreliable), validated against a typed schema, and cited back to a source page — confidence surfaced inline as evidence tags. Read the full write-up: Methodology · the research course · or about the method.
Pre-launch. Assembled from public records the developer didn’t write.
Registered-agent and organizer overlaps are common-control plumbing, not a
verdict on ownership; inference is labeled [inference] and never dressed
as fact. Nothing here is legal advice or a verdict — verify every figure against its
cited source.