Google has won a bankruptcy auction for part of Spirit Airlines’ enterprise data and software assets, offering $10 million for an unusually complete record of how a company operated over decades.
The deal is not final. Bankruptcy Judge Sean Lane must approve it at an August 19 hearing. But the winning bid reveals what has become valuable in the race to build workplace AI: not merely documents, but the connections among conversations, code, decisions and outcomes.
An airline captured in data
The proposed purchase includes about 100 million emails, 500 million Microsoft Teams chats and collaboration records, and 30 million lines of code with related development material. It also covers pricing for 7.2 billion competitor flights, roughly 7.5 billion passenger transactions dating to 2008, and more than 175,000 employee records reaching back to 1986.
Across those systems sit traces of aircraft operations, booking patterns, revenue management, finance, fraud investigations, marketing, human resources and corporate strategy.
Counting only the emails and Teams records, Google’s bid amounts to roughly 1.7 cents per communication. That assigns no value at all to the code, algorithms or operational data.
The archive is valuable because those records remain connected. An isolated report can describe what happened. A Teams discussion linked to a project ticket, code change and later business result can show how it happened. That sequence is particularly useful for AI agents meant to understand a workplace, navigate its systems and perform multistep tasks.
Why Google wants the links
Google says the acquisition could help improve its products and AI models. That fits the strategy CEO Sundar Pichai outlined in April:
“Enterprise data has become critical for agents to reason.”
An AI assistant can draft generic prose without knowing much about a company. An effective enterprise agent needs the company’s own context: who approved what, how internal systems relate, which process produced a result and what happened when workers made a particular choice.
Spirit’s archive offers that context at rare scale. It follows one organization across years of communications, software development and operating decisions, rather than presenting disconnected snapshots gathered from many places.
The price reflects that difference. Specialists in selling data from defunct companies have reported typical recoveries of $10,000 to $100,000. Google’s bid is 100 to 1,000 times larger, suggesting that Spirit’s long-running, cross-linked record is the asset, not simply the volume of text.
Deidentified does not mean uncomplicated
The sale excludes identifiable customer and privileged information, including 97.5 million passenger profiles and about 50.2 million Free Spirit loyalty records. Google says a third party will rigorously remove personally identifiable information before delivery, and the transaction prohibits attempts to re-identify customers.
The filings and available reporting do not identify that third party or explain the technical method. That leaves a consequential question around the included employee communications and HR material: how effectively can decades of workplace history be stripped of identity while preserving the relationships that make it useful?
This is a governance and engineering risk, not evidence that Spirit’s data will leak. But researchers have shown that language models can memorize and reproduce rare training examples, including conversations, code and personal identifiers. The more detailed and distinctive a record is, the harder it can be to treat deidentification as a simple deletion exercise.
Spirit stopped flying on May 2 during its second Chapter 11 case, laid off about 17,000 workers and entered liquidation. Its remaining corporate memory is now being priced as an AI asset.
That is the larger shift exposed by the auction. When a company fails, its desks, aircraft and software licenses are no longer the only things creditors can sell. The accumulated record of how its people worked together may be worth millions too.