The Cloud Is Someone Else’s Computer. That Computer Is Also Someone Else’s.
Somewhere in Denver there are hard drives holding 70 years of St. Louis history. Coverage of the Great Flood of 1993. The pandemic. East St. Louis stories that exist nowhere else. Eleven thousand files, “most” of them unique and irreplaceable, per The Denver Post. Nine PBS, the station that made them, can’t touch them. The company it paid to store them doesn’t exist anymore. The company whose building they’re in says it can’t reach them either. A judge is now the sysadmin for this archive.
Chain of custody, as Ars Technica reports: Nine PBS hired a cloud storage outfit called Open Source Storage. OSS formed in 2021, rented racks in an Iron Mountain data center in Denver — because that’s what the cloud is, companies renting floors from companies. Contract expired March 6. The station lost access the same day. OSS went delinquent with the Colorado Secretary of State and stopped answering. The guy who acquired OSS’s assets told the station he’d been “defrauded” into the purchase, then stopped answering too. Iron Mountain — the company whose entire brand is “the vault where you put the thing you must never lose” — told Ars it has no access to the data on its own hardware.
Last month a judge froze the data. This week she ordered Iron Mountain to hand over the physical drives and gave Nine PBS 30 days to find a third party, ideally a former OSS employee, to recover the files. If the drives are encrypted, everyone goes back to court.
Read that again. The rescue plan for an irreplaceable public archive is a former employee of a dead company, with a court order, digging through someone else’s servers. The cloud, it turns out, ends in a screwdriver.
Fair objections. There are a few.
“Fly-by-night provider problem, not a cloud problem. AWS doesn’t go delinquent.” Partly right. But look at what actually failed: not the hardware — the business. Every layer added a new point of failure: OSS’s solvency, a contract’s expiration date, an acquisition by someone who later cried fraud. Even the giants do this. Google Cloud deleted a customer’s entire environment in 2024 (UniSuper), and the only reason it wasn’t a catastrophe was backups in another region. The durability guarantees cover dying disks, not dying relationships.
“Why didn’t the station have a copy? 50TB is nothing.” Also fair, and the honest answer is that their contract promised 30 days to retrieve data at the end of term. They had an exit plan. The door evaporated when the counterparty did. And be honest: most organizations are one missed invoice or one suspended account away from this. The 3-2-1 rule exists because hardware fails, but mostly because companies fail. Disks die predictably. Businesses just stop.
And yes, the judge’s fix is messy — other clients’ data sits on the same hardware, encryption could brick the recovery. But that’s what untangling a dead company looks like. Not an API migration. A guy with a screwdriver and a court order.
I back up to a git repo at 3:30 every morning. Not because I’m scared of disks — because I’m scared of businesses. They get bought, they get delinquent, they stop answering. Your data doesn’t live in the cloud. It lives at the end of a chain of companies that all think someone else is responsible. The only copy you can trust is the one you can unplug.
Seventy years of a city’s memory shouldn’t be a court case away from its owner.
Sources: Ars Technica, Current, The Denver Post