Google buys collapsed Spirit Airlines' data at auction for $10M

Google has acquired a large trove of data belonging to Spirit Airlines, the US low-cost carrier that stopped flying for good in May 2026 and is now being wound up in liquidation. A court document filed last week shows that the airline's de-identified data was among the assets put up for sale, and that Google won it with a $10 million bid, subject to judicial approval.

What $10 million bought

The inventory is enormous: 100 million emails and 500 million items from Microsoft Teams, 17 million OneDrive files and 20.5 million items from SharePoint. It also includes more than 30 million recorded customer service calls and over 15 million customer service chat records. On the enterprise side sit 600,000 ServiceNow tickets, 13.7 million active email addresses taken from Oracle's Responsys marketing application, and details of 11 million purchases of in-flight Wi-Fi.

Operational records form a second layer: data describing more than 763,000 flights, five million crew pairings, upwards of 1.2 million fuel slips and records covering 787,452 parts purchases.

Why an AI company wanted an airline's archive

Google has reportedly said the purchase is meant to improve its AI services. The underbidder was Mercor, a company that supplies data for training AI models, which suggests the value of the trove is not unique to one buyer. The Register has reported that AI experts increasingly regard large language models as blunt instruments and see smaller models trained on a specific field of knowledge as more useful in some applications. Spirit's records could underpin an aviation operations model, or simply serve as a large body of ordinary business correspondence.

De-identified, but not risk-free

The court filing says the data was de-identified before it was offered for sale, and Google has promised to scrub any personally identifiable information it finds. Passengers who flew Spirit and worry that their flight history or a tense call with the airline's contact centre will end up in a model's memory are being told there is no need to be concerned. The practical history of such datasets, however, is that fragments of personal information can resurface in later searches or outputs.