Google's $10 million Spirit Airlines purchase is a bet that bankrupt companies are data goldmines

Google acquired data from bankrupt Spirit Airlines for $10 million, with no interest in the carrier’s aircraft or routes. Gonen Stein, co-founder and president of Eon, flagged the deal as evidence that companies in bankruptcy are becoming targets of a new kind of buyer: one that sees the accumulated records of daily commercial life as training material for artificial intelligence models.

The contents of what Google purchased clarify why $10 million went to a defunct airline rather than an aerospace supplier. Kevin Roose, technology columnist for The New York Times, reports that the deal included 100 million emails, 500 million Microsoft Teams chats and other conversations, 7.5 billion passenger transaction records dating back to 2008, and 30 million lines of Spirit’s internal source code. That is not an archive of flight schedules. It is a cross-section of how a mid-size American corporation communicated internally, processed transactions at scale, and built software over nearly two decades.

The volume and variety of that dataset matter because AI training does not improve on a single dimension. Language models learn from text in many registers: customer-facing, internal, technical, operational. A corpus that mixes executive email threads, customer service chats, transaction logs, and source code exposes a model to the kind of heterogeneous real-world language that synthetic data and public web crawls do not easily replicate. Spirit, as a mass-market carrier that processed billions of bookings over 17 years, generated that kind of breadth almost incidentally.

They didn't buy airplanes. They bought the data. They bought the data for $10 million because they think it's very important in that perspective. They're using that to train models. Gonen Stein

Stein’s framing places the transaction inside a larger argument about where valuable data actually sits. Legacy businesses, especially those that operated at volume for years before going under, often hold more raw transactional and communicative data than younger, leaner firms. Bankruptcy, paradoxically, makes that data accessible: the assets must be sold, and acquirers who would not have approached a functioning airline can bid on its records without needing to absorb its liabilities or workforce.

The scale of Google’s commitment to AI infrastructure adds context to why a $10 million data purchase is worth flagging at all. Robin Wigglesworth, editor of FT Alphaville at the Financial Times, notes that Google sits on roughly $800 billion in purchase commitments, with around $200 billion of that classified as short-term. Against those figures, $10 million is a rounding error. The significance is not the price but the category: Google treating a bankrupt airline’s records as a line item in its AI buildout suggests that data sourcing has moved well beyond licensing deals with publishers or partnerships with established data brokers.

Barry Knapp, managing partner of Ironsides Macroeconomics, puts the broader capital picture in sharper relief. He notes that the four largest AI capital-expenditure spenders are currently at 90 percent capex-to-cash-flow, above the prior peaks of roughly 80 percent seen in telecom during 1999 and 2000 and in energy during 2014 and 2015. That ratio signals the degree to which the largest technology companies have committed their near-term financial capacity to building AI infrastructure. At that level of commitment, every available data source acquires strategic value, including the records of companies that no longer exist.

What the Spirit Airlines deal illustrates is a market dynamic that has no obvious ceiling yet. Businesses fail constantly, and failed businesses leave behind data. The question receivers and bankruptcy courts have historically asked is whether that data has monetary value. Google’s $10 million answer, for a carrier most travelers had written off before its collapse, suggests the answer is yes more often than previously assumed. For institutions holding large stores of operational data, the implications extend well beyond aviation. The asset class, if it can be called that, is only now being priced.