When a startup closes, its code, emails, chats, and documents can become assets in a growing market. Companies that manage shutdowns are now helping founders sell those archives to AI labs and other buyers, who want material for training models and simulated work environments.
Shutdown services find a new asset to sell
Keelan Miskell did not expect much value from the records of his crypto startup. Its users were anonymous, and the company had operated for only a few years. But after deciding to wind it down, he sold its code base for roughly $25,000 and was awaiting a likely amount of around $15,000 for workspace data stored in Google, Notion, Figma, and Slack.
The deal went through SimpleClosure, a company that helps founders handle the practical work of closing a business. It added data sales to its services in April. Its Asset Hub platform connects founders with buyers for code and workspace records such as emails, chats, and documents.
Those buyers include foundation AI labs, companies that build simulated workspaces for AI training, and established tech companies developing their own models. Asset Hub has handled over 150 deals and expects to return over $10 million to startups by the end of 2026. CEO Dori Yona says 47% of SimpleClosure deals now include code or data sales.
Sunset is also building a resale business. Its founder and CEO, Brendan Mahony, says the company assesses assets using factors such as head count, years in operation, data volume, and how the team used its tools. That approach reflects a central idea behind the market: routine work records may have value even when they were never created as a product.
Why AI buyers want real company records
Interest in defunct companies’ data grew suddenly last year, as AI training labs encountered limits in public and synthetic datasets. Real workplace archives can show how people handle varied situations, including disagreements and difficult decisions. For AI buyers, that material may help build training environments where agents practice tasks.
Surge AI, for example, buys data to create “reinforcement-learning gyms.” Clients’ AI agents use these environments to practice activities such as debugging a complex IT architecture or drafting appellate briefs. Nick Heiner, who leads the company’s work on these environments, looks for archives that contain a range of situations and judgment calls.
The quantity of files alone does not determine an archive’s value. Heiner says one company’s Slack history may be more useful than another company’s Slack, email, and calendars combined if its conversations have greater variety and depth. Buyers are therefore evaluating not just how much data exists, but what kinds of interactions it captures.
Real archives can also avoid some errors found in synthetic environments, such as references to dates that do not exist. But the scale of any performance gains remains unclear: no company has released tests comparing training on real-world data with the alternatives. Heiner says buyers offer widely different prices because their valuations rest on research theses about how many useful tasks an archive might support.
Value comes with privacy and consent questions
A large deal involving Spirit Airlines illustrates the possible scale. In August, as part of the airline’s bankruptcy, Google outbid the AI data company Mercor to buy decades of company data, including 100 million emails, 500 million Teams messages, and 176,000 employee records. The reported price was $10 million. Bloomberg reported in September that SpaceX’s AI unit was weighing a purchase of data from struggling startups.
These sales raise concerns about personal information buried in ordinary business records. Spirit’s flight attendants’ union objected to the sale on privacy grounds. Greg Ewing, an attorney who advises AI startups, has called such data “legally radioactive.” Emails and calendars may contain details about medical appointments, school pickups, and personal plans, even when a company archive is treated as a business asset.
Suppliers say they use data scientists to remove personally identifiable information and exclude HR records, and buyers say they also scrub data. Still, Ewing describes a market only beginning to move from a “Wild West of AI training” toward licensing agreements and vetted use. The current process leaves questions about what is removed and how the remaining material may be used.
There is also a question of who benefits financially. Proceeds typically go to senior creditors and investors, rather than employees whose work helped generate the records. Ifeoma Ajunwa, a law professor at Emory University and author of The Quantified Worker, calls this a form of “captured capital.” She says employers should have express consent before selling workers’ data, while noting that there is no federal law banning such sales.
A market with uncertain terms
For a startup that is closing, selling data may recover some value from years of work. But founders have little to no visibility into who buys the archives or what the buyers intend to do with them. They also have little assurance that an offer reflects a fair price.
Buyers and sellers are still working out how to value the material, and the lack of published performance comparisons makes those estimates difficult to confirm. Data prices may stabilize in the coming year, unless an AI bubble pops first. For now, shutdown firms are appraising more than code and equipment: they are also finding buyers for the digital traces of how a company worked.