Spirit Airlines ceased flying on May 2, 2026, but the carrier's digital exhaust — millions of internal emails, chat messages, and lines of code accumulated over four decades — has become the subject of a multi-million dollar bidding war between Google and two AI startups.

The intensity of the bidding reflects a broader shift in the AI industry: as publicly available web data has become heavily scraped and contested, developers of large language models increasingly prize large, coherent corpora of real-world organizational communications — the kind of material that is difficult to synthesize and rarely available for licensing.

Google won the initial auction on August 14 with a $10 million bid for access to Spirit's de-identified business records. AI training startup Micro1 subsequently submitted a late counter-offer of $12.5 million, throwing the bankruptcy court's approval process into uncertainty ahead of a key hearing scheduled for September 9.

What is actually for sale

The dataset comprises roughly 100 million de-identified emails, approximately 500 million Microsoft Teams messages, around 30 million lines of custom software code, and employee files stretching back to 1986.

The assets explicitly exclude passenger profiles, loyalty program data, and personally identifiable information. This is an operational dataset — the type of internal communications corpus that reveals how a large organization functions day to day.

The bidding war

The auction unfolded in stages. Mercor, a data and talent company specializing in AI, served as the backup bidder with an offer of $7.5 million. Google topped that with $10 million and emerged as the winner.

The situation then grew complicated. Micro1, an AI training startup led by CEO Ali Ansari, submitted a $12.5 million counter-offer after the auction deadline had passed. That bid — 25% above Google's winning price — forced the bankruptcy court to weigh whether to reopen the process or uphold the original result.

The case is being overseen by Judge Sean H. Lane in the Southern District of New York. For context, Spirit filed for Chapter 11 bankruptcy protection a second time before halting operations entirely. Every additional dollar from the data sale goes toward paying down creditor claims; those creditors are unlikely to recover much from a defunct ultra-low-cost carrier.

The sale of corporate data assets out of bankruptcy estates remains a relatively novel frontier, and the ruling here could shape how courts, buyers, and employee representatives approach similar transactions involving the communications and records of defunct companies.

Union objections add another layer

The Association of Flight Attendants-CWA has filed objections raising serious concerns about privacy and the handling of employee data.

The union's argument is straightforward: even if the dataset has been stripped of passenger personally identifiable information, it still contains decades of employee files and internal communications. Flight attendants and other workers never consented to having their workplace communications sold to a tech giant for AI model training.

Google has indicated it would further de-identify the dataset before using it for AI model training, but the union's objections suggest that additional de-identification may not be sufficient to address their concerns.

The September 9 hearing will determine whether Google keeps its winning bid, whether Micro1's late offer receives consideration, and whether the union's privacy objections carry enough weight to block or modify the sale. Beyond the immediate outcome, the decision will be watched as an early test of how much say employees have over workplace data that outlives their employer.