What gives a data center away

Start with the people, not the code name.

I started by looking for secrecy: code names, NDAs and long tax abatements. That sounded reasonable until I checked the other big projects. Intel, Panasonic, Samsung, Scout and Foxconn used some of the same tactics. The more useful questions were who was behind the land deal and how many permanent jobs the project promised for the money going in. In this group of 38 projects, those two signals separated data centers from the other projects much better than the code names did.

The first rule missed almost everything.

My first pass added up seven signals, most of them related to secrecy. I set the rule before checking 15 more cases. It found only one of the nine data centers in that second group. I tried adding jobs per dollar and putting more weight on the companies in the filings. That helped. A simpler check did better in this sample: look for at least 100 acres and a known data-center company somewhere in the paperwork. I would not treat that result as a proven detection rate. Thirty-eight selected projects are too few for that.

The two persistent misses are Cleveland’s Slavic Village campus and the Nordson conversion in Amherst. The first was filed as a data center; the second was described that way at its first hearing. A method for spotting an undisclosed use has little to add when the use is already public. These misses still matter when you read the chart as a test of the whole sample.

The first filing can buy you time.

I could find both a first public trace and a later disclosure date for 13 Midwest projects. The gap was 14 months at the median, though it varied a lot. Toledo was the longest: a shell company sought rezoning in late 2022, “data storage” appeared in a zoning amendment nine months later, and Meta’s name surfaced 28 months after the first filing. If you begin with the hearing agenda, you may miss the earlier trail in deeds and company records.

Some names turn up again.

The company names change, but a few people appear in more than one filing. Indiana and Missouri records connect two managers to several LLCs associated with Google projects. The same lawyer appears on both Indiana filings in this set. That gives you names to search when a new LLC buys land. It does not prove who owns a new project; you still need the deed, the application or another direct link.

The same proposal lands differently in different places.

The regional count is uneven. Across four Northeast Ohio metros, I counted 15 proposals since 2022; seven had been withdrawn, rejected or stalled at the time of the sweep. I also found 30 local moratoriums in those metro counties. Pittsburgh’s area had seven proposals, one in the withdrawn or stalled group and one local moratorium in this count. Those numbers describe the records I found, not a vote on whether either region wants data centers. Zoning and site history matter too.

Where I would look next

Nine more records worth checking