The corpus Google acquired isn’t customer data — Spirit’s 97.5 million passenger profiles and 50 million loyalty records reportedly stayed out of the deal — it’s the digital exhaust of running an airline: emails, Teams chats, code, revenue and productivity records. That a bankruptcy court docket in the Southern District of New York, rather than a corporate biz-dev team, is now the venue where a trillion-dollar AI buyer sources training data says something about where the supply chain for proprietary text has moved.
At $10 million for roughly 600 million documents and messages, Google is paying a fraction of a cent per item — evidence that in this market, scale of corpus beats quality of curation as the price driver.
The more telling detail is who lost. Mercor.io, per Bloomberg’s reporting via SiliconANGLE, has spent recent months paying individual employees directly for materials from former employers — a Wall Street Journal report from April 2026 described the company canvassing professionals for Slack exports and internal documents to extract “industry-specific knowledge.” Losing a structured bankruptcy auction to Google suggests the retail model of buying data one ex-employee at a time doesn’t scale against a hyperscaler with cash and legal cover to buy an entire estate at once. Expect more AI labs to start treating bankruptcy dockets as a sourcing channel rather than relying on brokers.
The antitrust footnote nobody asked for
There’s an uncomfortable coda here for merger-review advocates. The Justice Department blocked JetBlue’s 2022 bid for Spirit specifically to preserve it as an independent low-cost competitor, a decision upheld by Judge William Young in January 2024. Bloomberg Law’s analysis notes roughly 17,000 jobs were lost and Spirit’s capacity vanished from the market anyway when it collapsed less than two years later. Now its remaining asset of value isn’t its route network or its brand — it’s the internal paper trail, sold off in pieces to the same class of AI companies critics say already dominate too much of the information economy.
What would change the read: how rigorously Google’s third-party PII scrubber actually filters 100 million emails and 500 million chat logs before training, and whether Spirit’s bankruptcy becomes a template other distressed carriers’ creditors start citing as a data-monetization option of last resort.
The haul reportedly spans 100 million company emails, 500 million Microsoft Teams chats, 30 million lines of code, software algorithms, development metadata and records pertaining to the airline's revenue, aircraft operations and employee productivity, Bloomberg reported.
— Reuters
