In August a US bankruptcy court took bids on an unusual asset. Not aircraft, not gates, not the loyalty programme, but Spirit Airlines' own internal records: roughly 100 million emails, around 500 million Teams messages, HR files and presentations, all deidentified.
Google bid $10 million and won it. Mercor, an AI hiring platform, was the underbidder at $7.5 million. It ran as a competitive auction, which is the only reason there is a price to talk about.
It's worth being clear about what changed hands, because a lot of the coverage wasn't. This was the company's internal corporate record rather than its passenger database. The flight attendants' union has objected to the sale, and that objection matters, because consent and provenance are now part of what any buyer has to work through.
Most of the open web has already been used as training material, and what the labs are short of now is an authentic record of how work actually gets done. How a decision got escalated, how a scheduling problem was argued out over three days, what a handover between shifts looks like when nobody is writing for an audience. You can't scrape that, because it was never public, and you can't synthesise it convincingly. A twenty year internal archive is one of the few places it exists.
That's one kind of buyer. There is a second kind, and they want almost the opposite thing. AI buyers want a one-off corpus: large, messy, unpolished, bought once. Funds, data aggregators and market intelligence firms want something narrow and structured instead, things like prices, volumes, lead times and churn, delivered on a schedule so one month can be compared against the next. A single dump is close to useless to them. A clean monthly feed is not.
Most businesses can serve one of those two buyers reasonably well, and hardly any of them have looked at their own systems and worked out which one.
A bankruptcy is a forced sale, and that is the only reason anyone knows what Spirit's archive was worth. Everything gets listed, valued and sold, including the things nobody bothered to value while the business was still trading.
A going concern never runs that process. The data sits on the balance sheet as nothing, so the option to do something with it never gets priced and never gets used. A business eight years into the same accounting system and six years into the same CRM is holding a version of the same asset. Smaller, obviously, but with one advantage Spirit's archive doesn't have: it is still being added to every day, and a recurring feed is worth more than a final one.
Xferdata values a company's data, builds the auction pack, and runs the auction to buyers globally.
If you own the data — sign up and run the assessment. It takes minutes and tells you whether what you hold is worth taking to auction, and what it's likely to be worth.
If you buy data — whether you're training models or looking for recurring feeds, sign up and tell us what you're after. You'll see packs as they come to market.