Google Pays $10 Million for Spirit Airlines' Corporate Data to Train Its AI Models

Verified key facts

• Google agreed to pay $10 million for Spirit Airlines' corporate data, disclosed in a court filing late on Monday 18 August, per CNN and Forbes.
• Google outbid Mercor.io Corp, an AI-focused recruitment firm, which offered $7.5 million.
• IT Pro reported the haul as roughly 100 million emails, 500 million Microsoft Teams chats and 30 million lines of code, alongside flight pricing records and internal documents.
• The purchase also covers Spirit's internally developed software: source code, plugins, data files, libraries, APIs and documentation.
• According to the court filing the data has been stripped of information that would identify individuals, and Google has said it will scrub personal information; passenger profiles and loyalty records were excluded from the sale.

An airline's filing cabinet, sold for the price of a house

Google has agreed to buy the corporate data of Spirit Airlines for $10 million, according to a court filing disclosed late on Monday and reported by CNN and Forbes. The sale covers the material an airline generates simply by existing: internal email, staff chat logs, spreadsheets, operational records, booking transactions and the software the company wrote for itself.

The sum is the striking part. Ten million dollars is a rounding error against Google's capital spending on artificial intelligence, and it buys a complete, coherent, decades-deep record of how one large operating business actually worked.

What is actually in the lot

IT Pro put the contents at roughly 100 million emails, 500 million Microsoft Teams chats and 30 million lines of code, alongside billions of flight pricing records and internal documentation. Tom's Hardware described the same lot in similar terms. Reported figures differ between outlets, which is normal for a filing summarised at speed, but the shape is consistent.

Google's money also buys Spirit's internally developed software outright: source code, plugins, data files, libraries, APIs and the documentation that explains them. That last category is the one engineers would notice. Documented code, paired with the chat logs of the people who wrote it, is a far richer training signal than code scraped without context.

Why an airline's chat logs are worth anything

The public internet has been comprehensively mined. What remains scarce is the private, structured, task-specific record of professional work: how a pricing decision was argued, how an incident was escalated, what a competent operations team says to one another at three in the morning.

That is precisely what a defunct company's archive contains, and it is why an AI-focused recruitment firm was bidding against Google at all. Mercor.io offered $7.5 million for the same material, per the reporting. Two very different businesses valuing the same corporate exhaust tells you the category has a market.

The privacy question, and how it was answered

According to the court filing, the data has been stripped of anything that would allow individuals to be identified, and Google has said it will scrub personal information. Sensitive categories, including passenger profiles and loyalty programme records, were explicitly excluded from the sale.

Those safeguards are meaningful and worth stating plainly. They also sit alongside a harder fact: the people whose working lives are recorded in 100 million emails and 500 million chat messages were employees, not parties to this transaction. Anonymisation addresses identification. It does not address the fact that an archive created for one purpose has been sold for another.

Bankruptcy as a data pipeline

The mechanism here is the genuinely new part. This was a court-supervised auction, the ordinary machinery by which a failed company's assets are converted to cash for creditors. That machinery now treats an information archive as a saleable asset like an aircraft or a landing slot.

Every company that fails from here has a data estate, and every creditor has an incentive to realise its value. Employees and customers of a business that goes under have no seat at that auction, and the terms on which their data was originally collected were not written with one in mind.

There is a further wrinkle in who ends up holding it. Spirit's archive documents not only Spirit but everyone it dealt with: the airports it negotiated with, the lessors it leased from, the suppliers it argued with over invoices. Those counterparties never had a relationship with the buyer and had no standing in the auction, yet their side of years of correspondence is inside the lot. Commercial confidentiality is normally protected by contract between two parties; it is not obvious that those contracts survive one of the parties being liquidated and its filing cabinet sold to a third.

What it signals about the training-data market

The AI industry spent the last two years signing licensing deals with publishers, forums and stock libraries. This is a different transaction: not licensed content produced for an audience, but the operational residue of a company that never intended any of it to be read outside the building.

If $10 million is the clearing price for one mid-sized airline, the aggregate value locked up in the archives of every distressed business is substantial, and the legal apparatus for releasing it already exists. That is a supply channel the licensing debate has largely not addressed.

Whether creditors elsewhere take the hint

The immediate markers are procedural: whether the sale draws objections before it is finalised, and whether regulators take an interest in the scrubbing commitments rather than accepting them as given. Reuters reported that a US court delayed a hearing connected to the purchase, so the process is not yet closed.

The broader marker is imitation. If other estates in bankruptcy begin listing data as a separate asset class, this $10 million filing will be read as the transaction that set the price.

Sources

• CNN Business - Google is buying all of Spirit Airlines' data to feed its AI models
• Forbes - Google Paying $10 Million For Spirit Airlines' Data To Train AI
• IT Pro - Google spent $10 million in an auction for Spirit Airlines data
• Tom's Hardware - Google buys Spirit Airlines data for AI training
• Reuters - US court delays hearing on Google's purchase of Spirit Airlines data