Spirit Airlines, the budget carrier that failed earlier this year, just sold its old files to an unlikely buyer.
The purchase happened inside Spirit's Chapter 11 case, the court process companies use to restructure debts when they can no longer pay what they owe. Court documents confirm Google was the winning bidder.
A Google spokesperson said the material "can be helpful in improving our products and AI models." The company also made clear what it did not buy: "We are buying the company's internal data and custom software, but we are not buying their customer or credit card information."
No passenger details and no credit card numbers changed hands. In other words, Google gets the airline's operational history, not its customers' personal lives.
The material includes the airline's own records and the custom software the airline built to run parts of its operation. It is the kind of digital leftover most companies never think about selling.
Spirit entered bankruptcy after years of mounting losses and a failed merger attempt, leaving creditors to claim whatever assets remained. In this case, the most valuable asset turned out to be the company's accumulated data and proprietary systems - a stark reminder that even failed businesses can hold hidden worth.
This was not a quiet sale, and bidding for Spirit's data was competitive, according to court documents. It started with Google at $5 million, then Mercor, an AI-training firm, countered at $5.2 million. Mercor later proposed $7 million, but only if it received the data before any identifying details were removed. Google came back with a final $10 million bid, which beat Mercor's separate $7.5 million offer.
The price jump from $5 million to $10 million shows how badly AI companies want raw material. A bankrupt airline's old records are now a valuable asset.
Mercor did not reply when Business Insider asked for comment, and Spirit gave no statement beyond its court filings.
The fight over Spirit's files is part of a larger shortage. Tech companies have trained AI on the open internet for years, and most of that free public data has been used up.
Now they are looking for fresh material in private hands. Corporate records are attractive because they come with real-world problems built in.
A customer complaint log shows how people actually speak and what actually goes wrong, which is exactly what an AI needs to learn if it is going to handle similar situations. Website bug reports and internal software teach the same kind of lessons.
That practical experience is hard to find in the polished text and images that fill the public internet. The market for this material already exists.
Anonymized data is the kind that can change hands without exposing real people. Names, contact details, and other identifying markers are removed, which makes the files safer to sell.
Micro1, an AI-training startup, said earlier this year that it pays 50 midsize companies between $100,000 and $2 million for anonymized data, meaning data with personal details stripped out.
The race is also producing some strange moves. Meta, the owner of Facebook, tried monitoring employees' keystrokes and mouse movements to train an AI on how to use a computer.
The company suspended the effort after an internal data exposure and employee backlash.
For investors, this deal is a peek into where the AI boom is headed. Big tech has run low on free public material, and it is now writing checks for the messy data locked inside private companies.
The value of a business used to come from its products, its customers, and its brand. Increasingly, it may also come from the records it keeps, because someone out there is willing to pay for them.
The deal also shows that the AI industry will go where the data is, no matter how odd the source. A budget airline that failed this year just sold its digital history for $10 million.
Data has become an asset with real dollar value, and the companies sitting on it may be worth more than their earnings alone suggest.
