Google Just Won an Auction for 100 Million Spirit Airlines Emails. What It Really Bought Was a Company’s Memory

Google has won a $10 million auction for a massive archive of Spirit Airlines’ internal business data, including around 100 million emails and 500 million Microsoft Teams messages. The company says the material could help improve its AI. The sale still needs court approval.

Spirit Airlines Airbus A320 departing Baltimore-Washington International Airport. Photo: Acroterion / Wikimedia Commons, CC BY-SA 4.0.

Spirit Airlines stopped flying months ago, but pieces of the company are still being sold. Its planes have value. Its airport slots have value. Its headquarters had value. And apparently, so do hundreds of millions of old workplace conversations.

Google has won a bankruptcy auction with a $10 million bid for a huge collection of Spirit Airlines’ internal business data and custom software. According to Reuters, Google says the material could eventually be used to improve its products and AI models. The sale is not completely finished yet; a U.S. bankruptcy judge is scheduled to consider approving it on August 19.

The archive reportedly contains roughly 100 million emails, 500 million Microsoft Teams messages, calendars, spreadsheets, presentations, marketing information, HR and productivity records, finance and accounting material, operations databases, pricing models, booking information, refund records and custom software. This is not just a pile of old files. It is a record of how thousands of people actually ran a large company, and that may be exactly why AI companies want it.

Google is not buying Spirit passengers’ personal data

Before going any further, there is an important distinction. Google is not buying Spirit’s customer or credit-card information. Reuters reports that personally identifiable information is supposed to be removed before Google receives the dataset. A third party is expected to de-identify the material, and Google has said it will not receive passengers’ personal information.

Axios also reports that passenger profiles and frequent-flyer information are outside the Google transaction. So the story is not “Google bought the personal data of everyone who ever flew Spirit.” What Google is buying is arguably stranger: the internal history of how Spirit Airlines worked.

Why would Google want 500 million Teams messages?

Modern AI models have already absorbed enormous amounts of public information: books, websites, Wikipedia, code, academic papers, forums, news articles and social media. But public information usually shows the finished version of something. Corporate records show the messy process behind it.

Imagine a company has a serious operational problem. One employee notices it. Another investigates. A manager gets involved. Finance questions the cost. Operations proposes a solution. Someone else says it will not work. A new plan is created, that plan fails, and eventually another team fixes it. A public article might summarize all of that as: “Spirit Airlines experienced an operational disruption.” Internal emails and Teams messages contain the actual reasoning.

Google told Business Insider that the dataset could be used to improve its products and AI models. This is the part I find most interesting. For years, AI companies raced to collect what humans published online. Now the valuable material may increasingly be what humans wrote privately while actually doing their jobs.

Google headquarters. Photo: Anthony Quintano / Wikimedia Commons, CC BY 2.0.

Another AI company wanted the archive too

Google was not alone. AI company Mercor also bid for Spirit’s data and ended up as the backup buyer with an offer of $7.5 million. Reuters says Google ultimately won with its $10 million bid.

Business Insider reported that the auction became a genuine bidding competition. Google reportedly began around $5 million, Mercor responded with $5.2 million, and the bidding continued until Google jumped to $10 million.

Mercor’s explanation for wanting the archive is probably the easiest way to understand the entire story: real corporate records can show AI systems how actual work gets done. That means two AI companies were essentially competing to buy the digital memory of a failed airline. Ten years ago, that sentence would have sounded ridiculous. Today, it makes business sense.

What is actually inside Spirit’s corporate archive?

Skift reports that Spirit’s records include information connected to finance and accounting, revenue management, operations, pricing models, booking curves, flight behavior, inflight sales, Wi-Fi sales, refund histories, employee systems and travel-package operations.

The archive reportedly stretches back decades. Skift says it includes around 3.4 million payroll records, material connected with roughly 80,000 email accounts, and employee-related information going back as far as 1986. Other reporting based on the bankruptcy documents has put the amount of custom software involved at roughly 30 million lines of code.

So Google is not simply getting a mailbox export. It is potentially getting a historical model of how a large airline made decisions.

Failure might actually make the data more valuable

Spirit Airlines failed. Normally, that would make its business history sound less valuable. For AI training, the opposite could be true.

A successful company teaches you what worked. A failed company can also show you what did not work. Spirit dealt with aircraft operations, scheduling, pricing, customer complaints, staffing problems, competition, financial pressure, changing demand and eventually bankruptcy. Somewhere inside those hundreds of millions of messages are likely long chains of decisions followed by real consequences.

To be clear, Google has not said exactly which models will use which parts of Spirit’s data. Claims that Google is building a dedicated “airline AI” would therefore be speculation. But possible uses could include AI agents that understand enterprise workflows, systems that reason across operational and financial information, models that follow long chains of workplace communication and AI tools that identify why business processes fail.

$10 million suddenly sounds cheap

Just look at the scale: 100 million emails plus 500 million Teams messages equals around 600 million pieces of internal communication, before counting spreadsheets, calendars, presentations, databases and software.

Obviously, messages cannot simply be valued by dividing $10 million by 600 million. Some will be useless, repetitive or irrelevant. But it illustrates something important: corporate data that used to sit forgotten on servers may now have significant value because AI developers want examples of how real organizations operate.

A company can disappear. Its data can keep appreciating.

The employee question is harder

The customer-data issue is relatively clear because personal passenger information is supposed to be excluded or removed. The employee side feels more complicated.

Former Spirit employees did not write millions of internal messages while thinking, “Maybe Google will use this to improve an AI model one day.” They were doing their jobs: discussing schedules, solving problems, arguing about decisions, sending spreadsheets and reporting mistakes.

The records are supposed to be de-identified before Google receives them, which is important. But removing names does not remove the underlying human behavior contained inside the conversations. That could eventually become a much bigger debate.

A bankrupt company’s data is becoming an asset class

Spirit may simply be an early example of something we are going to see much more often. Companies going through bankruptcy have always sold whatever still has value: buildings, equipment, patents, brands, software, aircraft and customer lists.

Now add something new: the accumulated record of how employees worked.

Business Insider reports that AI data company Micro1 has already offered midsize companies between $100,000 and $2 million for access to anonymized operational information. That suggests an actual market is starting to form. AI companies are not only hunting for more data. They are hunting for better kinds of data.

Spirit itself makes the story even stranger

Since Spirit stopped operating, the company has been broken apart piece by piece. Skift reports that JetBlue bought 22 Spirit slots at New York’s LaGuardia Airport for $58.5 million, while Spirit’s Florida headquarters was sold separately for around $93 million. Now Google has offered $10 million for the corporate data.

There is something strangely symbolic about that. The planes have value because somebody can fly them. The airport slots have value because another airline can use them. The headquarters has value because another company can occupy it. The software has value because somebody can run it. And the emails have value because an AI can potentially learn from them.

There is still one important catch

Google has won the auction, but the transaction has not yet received final court approval. A federal bankruptcy judge is scheduled to consider the sale on August 19. If the Google transaction fails to close, Mercor remains the backup bidder at $7.5 million.

So the accurate story right now is: Google won the bidding for Spirit Airlines’ internal business archive. It is not accurate to say Google already owns every piece of Spirit’s data, and it is definitely not accurate to say Google bought every Spirit passenger’s personal information.

The real story is bigger than Spirit Airlines

At first glance, this looks like a weird bankruptcy story: Google pays $10 million for old emails. But I think something much bigger is happening.

For the first stage of the AI boom, companies scraped the internet. The next stage could involve AI companies competing for the private operational history of real businesses: how employees communicated, how decisions were made, how projects failed, how problems were solved and how organizations actually functioned.

That information is difficult to reproduce artificially because nobody originally created it for AI. It exists because people were simply working.

Spirit Airlines may never fly another passenger, but decades of its decisions, arguments, mistakes, software and workplace conversations could still contribute to the next generation of artificial intelligence.

And I genuinely cannot decide whether that is brilliant, unsettling, or probably both.