Notice
This article is based on publicly available materials from Spirit Aviation Holdings filed with the U.S. Bankruptcy Court for the Southern District of New York, as well as materials from Google, Reuters, the Association of Flight Attendants-CWA (AFA-CWA), Mercor, and other sources available as of August 19, 2026. DANA NOTES’ analysis is also included.
What This Article Covers
Google has bid $10 million for the enterprise data of U.S. airline Spirit Airlines, which is undergoing bankruptcy proceedings.
In the auction held on August 14, 2026, Google was selected as the highest and best bidder, while AI company Mercor bid $7.5 million and was selected as the backup bidder. The transaction requires court approval, and the approval hearing was postponed to September 9, 2026, following an objection from the union representing Spirit flight attendants.
At first glance, the scale of the transaction draws attention to 100 million emails and 500 million Microsoft Teams data records. It also includes records related to flight operations, crew scheduling, fuel, maintenance, pricing, revenue, refunds, accounting, and software development. Google said it plans to use the data for product development and AI model training.
This case shows the economic value that data accumulated through a company’s work can have in the AI era.
Key Terms to Know
De-identified Data
Data from which information that can directly identify a specific individual, such as a name or contact information, has been removed or transformed to make it difficult to link the data to that individual. In this transaction, the contract requires the data to undergo a separate de-identification process before it is transferred to Google.
Domain Data
Data accumulated through actual work and operations in a specific industry, such as aviation, finance, or manufacturing. It can contain industry-specific workflows, system usage patterns, recurring work situations, and operational patterns.
Referential Integrity
The preservation of relationships when different data records are linked to the same person, task, event, or item. Spirit’s sale agreement requires the referential relationships across the entire dataset to be preserved even during the de-identification process.
The Enterprise Data Google Bid $10 Million For
Spirit Airlines’ data auction was held on August 14, 2026.
According to court documents, Google bid $10 million and was selected as the highest and best bidder, while Mercor.io Corporation was selected as the backup bidder with a bid of $7.5 million.
At this point, it is more accurate to view the $10 million as the winning bid amount pending court approval rather than the final completed transaction amount.
AFA-CWA, which represents Spirit flight attendants, objected to the way employee data would be handled, and the approval hearing was postponed to September 9. AFA-CWA has also stated that the data has not yet been transferred to Google and that approval of the transaction remains pending.
Google explained that the transaction does not involve purchasing Spirit’s customer lists or credit card information and that the information it receives will undergo a third-party de-identification process.
What Data Is Included?
The assets for sale include a wide range of data generated through the actual operation of an airline.
The following employee work records are included:
- Approximately 100 million emails
- Approximately 80,000 email accounts
- Approximately 500 million Microsoft Teams data records
- Approximately 17.08 million OneDrive records
- Approximately 20.58 million SharePoint records
- Approximately 668,000 ServiceNow IT tickets
Software development data is also included.
- 516 web, mobile, and backend source code repositories
- Approximately 30 million lines of source code
- Approximately 43,000 Pull Requests
- Approximately 370,000 Git Commits
- Bugs, issues, and related discussion records
The dataset therefore includes everything from documents and communication records created by employees to records of how actual software was developed and modified.
Actual Airline Operational Records Are Also Included
The scope of Spirit’s data extends beyond office documents and messages.
It also includes various business system data generated through airline operations.
- Approximately 763,000 actual flights
- Approximately 5.01 million crew pairings
- Approximately 787,000 maintenance parts receipts
- Approximately 1.239 million fuel-related records
- Approximately 3.53 billion Farebasis observations related to fares
- Approximately 7.25 billion competitor flight price observations
- Approximately 190 million PNRs for reservation confirmations
- Approximately 7.5 billion revenue system transaction records
- Approximately 9.29 million refunds
- Approximately 18.24 million Credit Shells
- Approximately 12.14 million Vouchers
In the finance and accounting area, supplier information, invoices, cost accounting, financial models, budget-related materials, and board financial reporting materials are listed as included assets.
Viewed as a single enterprise operating structure, this data records processes across multiple systems: planning flights, assigning crews, managing fuel and parts, setting ticket prices, processing reservations and payments, and completing refunds and accounting.
From an AI perspective, it becomes data that can show not only a specific outcome but also the work process that produced that outcome.
Customer Service Calls and Customer Profiles Are Excluded
The scope of the data in this transaction also needs to be distinguished precisely.
In the Assets Schedule submitted to the court, a substantial amount of customer-related data is marked Not Included.
The following data, among others, is excluded:
- Approximately 97.5 million customer profiles
- Approximately 50.2 million Free Spirit members
- Approximately 13.7 million active customer email addresses
- Approximately 15.78 million customer chats
- Approximately 30.87 million customer service call recordings
- Approximately 7.34 million phone numbers
- Website usage, search, and purchase analytics data
Therefore, the approximately 30 million customer service call recordings mentioned in some early reports are not included in the data Google is acquiring.
By contrast, operational data generated in enterprise systems, including reservations, pricing, revenue, and refunds, is listed as included. This data must also meet the de-identification requirements specified in the contract before it can be transferred.
Why Does Google Want This Data?
Google explained that Spirit’s enterprise data could help improve its products and AI models. Reuters reported that Google plans to use the data for product development and AI model training.
The defining characteristic of this data is that it consists of work records generated inside an actual company.
For example, publicly available aviation information on the internet can provide information such as flight schedules, ticket prices, and aircraft types.
Spirit’s enterprise data can show a broader flow of work.
It contains records of what materials employees shared, how projects progressed, what tasks were processed in systems, what problems software went through before being modified, and what data was used to manage flight operations and pricing decisions.
As AI expands toward finding documents within companies, analyzing work, modifying software, and performing tasks across multiple systems, the potential uses of this type of work data can also expand.
The Business Question to Watch
The important question in this transaction is:
How large a data asset can the work processes accumulated by a company become?
Companies generate large amounts of data every day.
Employees send emails, hold meetings, create documents, enter transactions into ERP systems, record customer interactions in CRM systems, modify code, and process costs and revenue in accounting systems.
Each record is originally created for its own business purpose.
As AI expands as a tool for learning and supporting enterprise work, these records are gaining a second potential use. They can be used as training and evaluation data that shows how actual work is performed.
In Spirit’s case, an actual bid price has been placed on that potential.
Google bid $10 million, and Mercor also bid $7.5 million. The fact that two different AI-related companies offered prices in the millions of dollars for the same enterprise data asset demonstrates the economic value of enterprise work data.
The Connections Between Work Processes Are Also Becoming Important
The value of enterprise data is difficult to explain by the number of files alone.
Spirit’s data contains records generated through different types of work.
For example, software development can produce a flow like this:
Work Request → Development Issue → Code Change → Pull Request → Review → Deployment Record
Airline operations can produce another flow:
Flight Planning → Crew Assignment → Maintenance → Fuel → Actual Flight → Revenue·Refunds
Corporate management can create another set of connections:
Business Plan → Budget → Supplier Transaction → Cost → Accounting → Management Reporting
Each dataset can provide information independently, but as the datasets become more connected, they can provide a richer view of how the company actually carried out its work.
Spirit’s sale agreement requires Referential Integrity, meaning the relationships between data records, to be preserved even after de-identification. This also reflects the value of this type of data structure.
A Market for Trading Enterprise Data Itself Is Also Emerging
There is no need to view the Spirit case only as a one-time data sale by a bankrupt company.
AI data companies are now also beginning businesses that de-identify internal work data from companies that continue to operate and provide it to AI companies.
Mercor officially offers a service that anonymizes operational data held by companies and licenses it to AI labs. The target data includes emails, messages, documents, project management records, code, and financial and operational systems.
Micro1 also operates a data partnership business that connects enterprise operational documents, CRM data, project records, and work decision-making processes so they can be used as AI training data. The company explains that compensation is calculated based on the quality, rarity, and usefulness of the data.
If this trend expands, three types of participants could emerge in the enterprise data market.
- Companies that hold data
- They can use data accumulated through their work processes as a new asset.
- Companies that process and intermediate data
- They handle personal and confidential information and organize data structures into forms that AI companies can use.
- Companies that develop AI models and services
- They can obtain data generated through actual work in specific industries and use it for model training and evaluation.
Spirit’s auction is a case in which this type of market structure has also appeared in a bankruptcy asset sale.
Companies Providing Data May Also Gain a New Revenue Source
From a company’s perspective, data has so far been used mainly for internal analysis and operations.
If AI training data transactions expand, companies may be able to organize the usable portions of the data they hold and connect them to licensing revenue.
The business models currently presented by Mercor and Micro1 are also moving in this direction. Companies select the scope of data they will share and provide it to AI companies after de-identification and data preparation.
Companies in industries such as manufacturing, finance, logistics, and aviation that have operated business systems for long periods have accumulated data containing industry-specific work patterns.
If an AI developer wanted to create the same kind of data directly, it would need to operate an actual business and accumulate data over a long period or go through separate simulation and data creation processes.
If AI developers can transact with companies that have already accumulated such data, they can obtain the industry data they need more quickly, while the companies providing the data can use assets generated through existing work processes as a new revenue source.
In this case, the price of data is likely to vary based on its scale as well as rarity, quality, time span, work-process connectivity, usage rights, and de-identification feasibility. Micro1 also identifies quality, rarity, and data value as compensation criteria for its enterprise data partnerships.
As Data Value Grows, Governance Becomes Part of Asset Management
As enterprise data gains economic value, the role of data governance also grows.
Spirit’s agreement requires a de-identification process before the data is transferred, with Google responsible for the cost. Google has also accepted conditions requiring it to keep the data in a de-identified state and not intentionally link it to a specific individual or household. If Google later transfers the de-identified data to a third party, the same conditions must also be imposed through contract.
This reveals an important characteristic of enterprise data assets.
A company possessing data and that data being in a condition that allows it to be sold or licensed are two different matters.
Companies need to manage conditions such as the following for each dataset:
- Which system generated it
- Who holds the rights to use the data
- Whether it contains personal information
- Whether it contains corporate confidential information
- Whether it can be provided externally
- Whether its business structure can be preserved after de-identification
- Whether AI training or reuse is permitted under the relevant contracts
The clearer this information is, the easier it becomes to determine how the data can be used.
De-identification and Data Connectivity Must Be Managed Together
This issue has become a point of dispute in the Spirit data sale.
AFA-CWA raised concerns that employee emails, Teams data, payroll records, and other employment-related materials are included in the assets being sold. The union argued that even if names are removed from the data, the confidentiality of the content itself is a separate issue, and it raised the possibility that information about specific individuals or small groups could be inferred again when relationships across multiple systems are preserved.
The agreement itself requires referential integrity across the entire dataset to be maintained during the de-identification process.
The use of enterprise data for AI requires both goals to be managed together.
Preserving the business value of the data while protecting individuals and sensitive information is the task.
How this balance is designed can determine the extent to which data can become an asset usable for AI.
The Scope of Data Due Diligence Could Expand in Bankruptcy and M&A
The Spirit case also provides a new perspective on corporate transactions.
When a company goes bankrupt or is acquired by another company, cash, equipment, real estate, intellectual property, contracts, and the customer base have traditionally been evaluated as important assets.
As separate market prices begin to form for enterprise data, data itself can become a more specific due diligence target.
For example, an acquirer may examine:
- What types of data have been accumulated and in what quantities
- How many years of work records are held
- Whether the data is connected across systems
- Whether rights for external use have been secured
- Whether personal information and confidential information can be separated
- Whether the data is in a condition that allows it to be used for AI training or analysis
Spirit’s highest bid of $10 million does not become a pricing benchmark for all enterprise data.
However, this case provides a real example in which enterprise data can be separated from other assets and receive a price as an independent transaction asset.
Competition for Enterprise Data Could Become Competition to Secure Industry-Specific Work Experience
Airlines have their own work structures.
Flight planning, crew scheduling, aircraft maintenance, fuel management, fare management, reservations, payments, and refunds are connected with one another.
Manufacturing has work structures involving production, quality, equipment, and supply chains, while finance has processes for transactions, underwriting, risk management, and regulatory compliance.
These industry-specific operational records become Domain Data.
As AI performs more actual work within specific industries, the potential value of work data accumulated in those industries can also increase.
This is why competition to secure enterprise data could move beyond collecting documents and become competition to secure actual industry-specific work experience in the form of data.
DANA NOTES Commentary
The most notable business change in the Spirit Airlines case is that managing data well could become directly connected to asset value in the future.
Companies already have large amounts of data. Going forward, an important difference may be created not by how much data they hold, but by whether they have organized it into a form that can be used when needed.
For example, even if a company holds decades of work records, the scope for external use may be limited if usage rights are unclear and personal and confidential information are mixed together. If the sources and rights relationships of the data are organized, the necessary portions can be de-identified, and the relationships between datasets are also managed, the company can have far more options for using the data.
This change could also broaden the nature of data governance. Data governance can serve as a management system for security and regulatory compliance while also becoming the foundation that determines how a company can use the data it holds as an asset.
If the data transaction market grows, new roles may also become necessary within companies. These include determining which data can be provided externally, separating personal and confidential information, confirming usage rights, and constructing valuable datasets. Like financial due diligence and intellectual property due diligence, this is an area where the importance of Data Due Diligence could increase.
Changes are also beginning on the market side. Companies such as Mercor and Micro1 have already emerged to connect enterprise operational data with AI companies and support de-identification and licensing.
If this structure expands, companies providing data can use data generated through their existing work as a new revenue source, while AI companies can obtain actual work data accumulated over years in specific industries within a shorter period. A new market can also emerge for companies that intermediate data, perform de-identification, and manage rights.
Spirit Airlines’ $10 million bid is one enterprise data transaction. At the same time, it can be viewed as a signal that work records accumulated inside companies are beginning to be valued as tradable business assets in the AI era.
Variables to Watch Going Forward
The first schedule to watch is the court’s sale approval hearing on September 9, 2026. The first variable is whether the court approves the current transaction terms and whether additional conditions are imposed on employee data and the de-identification process.
The second variable is the actual de-identification method. The agreement requires personal information to be removed while preserving the referential relationships between data records. How personal information protection and data usability are implemented together in the actual processing will be important.
The third variable is the final data transfer and the scope of Google’s use. Google is currently the highest bidder, and the data will be transferred after court approval and the contractual conditions are satisfied.
Finally, it will also be necessary to watch how much similar transactions increase. If more companies begin licensing enterprise operational data and data contracts between ordinary companies and AI companies become more common, enterprise data could become a new supply market for the AI industry. Mercor and Micro1 are already commercializing the licensing and monetization of enterprise operational data, making the formation of this market worth continuing to watch.
