Google secured $10 million in corporate data and custom software from Spirit Airlines' bankruptcy auction, outbidding AI training startup Mercor in a competitive process that reveals how scarce high-quality training material has become.
Mercor initially offered $7.5 million. Google opened at $5 million, Mercor countered at $5.2 million, then proposed $7 million if allowed to anonymize the raw data before processing. Google's final bid of $10 million won the auction. Court filings confirm the purchase excludes customer and credit card information.
The data is valued for training AI models to handle customer service inquiries and debug software—tasks that require understanding how real work operates inside an organization. A Google spokesperson confirmed the purchase is intended to "improve our products and AI models."
The bidding intensity reflects a tightening supply constraint. Technology companies have depleted most of the open internet as a training source and are now competing directly for proprietary datasets. Mercor's spokesperson stated the core thesis plainly: "Companies are sitting on decades of records that show how real work gets done, and that data is now some of the most valuable material for training and evaluating AI." Mercor operates a licensing program seeking operational data from major firms.
Other AI developers are pursuing similar strategies. Micro1 launched a program offering $100,000 to $2 million to 50 midsize companies for access to anonymized operational data.
Corporate data acquisition carries material risks. Meta initiated a program to track workers' keystrokes and mouse movements to improve its AI's computer-usage capabilities, but suspended it after an internal data exposure and employee backlash.
For Google, a $10 million investment in Spirit Airlines' data represents a modest outlay relative to its scale—suggesting marginal immediate value but strong perceived utility. The competitive bidding itself is the signal: when multiple AI firms are willing to bid against each other for a bankrupt airline's internal records, the market for proprietary datasets has become a structural constraint on model development.
