Google wins $10M Spirit Airlines data auction for AI training

Alphabet's Google won a bankruptcy auction for Spirit Airlines' internal business data, paying $10 million for a trove spanning 100 million emails and 500 million Microsoft Teams messages to train its AI models.

"We acquired part of an enterprise dataset from Spirit Airlines, which can be helpful in improving our products and AI models," a Google spokesperson said. "Any data we receive will be rigorously scrubbed of any personally identifiable information by a third party before receipt."

The dataset includes 30 million lines of code, software algorithms, development metadata, spreadsheets, calendars, marketing records, and documentation covering finance, accounting, website analytics, and aircraft operations. Google outbid AI data company Mercor, which submitted a $7.5 million final offer after earlier bids of $5.2 million and $7 million. Google's winning bid followed an initial $5 million offer.

The transaction, subject to approval at an August 19 hearing in the US Bankruptcy Court for the Southern District of New York, reflects a broader shift as AI developers exhaust publicly available internet data and turn to proprietary corporate records for industry-specific training material.

What's in the dataset

The acquisition excludes customer-facing data. Spirit's 97.5 million passenger profiles and more than 50 million loyalty program member records were not part of the sale, according to court documents. The airline retains the right to keep copies of the data for winding down its business and complying with legal obligations. Google confirmed it is not purchasing customer data or credit card information, and the data must be de-identified before transfer.

Spirit Airlines ceased operations in May after failing to emerge from its second Chapter 11 bankruptcy restructuring, weighed down by high debt and elevated fuel costs. The airline first filed for bankruptcy on Aug. 29, 2025. Attorneys and investment bankers, including PJT Partners, continue to manage the disposition of remaining assets.

The market for dead companies' data

The sale reflects a growing market in which technology companies, having largely exhausted publicly available internet data for training large language models, are turning to proprietary corporate datasets. Failed startups have increasingly sold internal communications and operational records to AI developers seeking industry-specific knowledge.

Earlier this year, AI training startup Micro1 announced a program paying 50 midsize companies between $100,000 and $2 million for access to anonymized data. Meta separately launched an initiative to track employee keystrokes and mouse movements to improve its AI systems, though the program was suspended following internal data exposure and employee backlash.

The competition for such data has intensified as researchers warn the world is running out of fresh training material. A Stanford report published in April flagged that AI developers face a looming data ceiling, with public web sources largely mined. That scarcity has pushed companies like OpenAI and Anthropic to pursue alternative data sources, including licensing deals with publishers and content platforms.

For Alphabet shareholders, the $10 million expenditure is immaterial relative to the company's capital allocation. But the deal shows how AI companies are competing for scarce training data as the industry confronts a potential data ceiling. Google's Gemini suite stands to benefit from the airline's operational records, which offer real-world business workflows that synthetic data cannot replicate. Analysts maintain a consensus Strong Buy rating on Alphabet stock, with an average price target implying roughly 23 percent upside from current levels.

This article is for informational purposes only and does not constitute investment advice.