When Keelan Miskell wound down his cryptocurrency startup this year, he assumed the company’s digital archives held little value. The user base was anonymous, and the firm had operated for only a few years. Instead of discarding the records, Miskell sold the code base for approximately $25,000. He is also awaiting a pending payment of roughly $15,000 for workspace data stored in Google, Notion, Figma, and Slack. “At least all that stuff we were doing for two and a half years is worth something to someone somewhere,” Miskell said.
Miskell’s transaction was facilitated by SimpleClosure, a firm that has expanded its services to include the sale of data from dying companies. This activity is part of a broader ecosystem supporting business failures. Estimates from Carta and AngelList indicate that startup shutdowns have risen between 26% and 56% annually in recent years. This growth has spawned a sector of “tech undertakers” that handle legal, tax, and operational wind-downs. SimpleClosure added data sales to its offerings in April, and CEO Dori Yona noted that these transactions have shifted from rare and taboo to routine. Forty-seven percent of SimpleClosure deals now include the sale of code or data.
The Market for Operational Exhaust
SimpleClosure’s Asset Hub platform has processed over 150 deals and projects returning more than $10 million to startups by the end of 2026. The buyers include foundation AI labs, vendors that build simulated workspaces for training, and mature technology companies developing their own models. The industry rival Sunset also operates a rapidly growing data resale business. Founder Brendan Mahony compares the data to wood shavings from particle board production. He argues that this “operational exhaust” contains significant value despite being previously overlooked.
The demand for this type of information emerged suddenly last year as AI training labs reached the limits of public and synthetic datasets. Surge AI, a company that purchases data to create reinforcement-learning environments, seeks archives that reflect a wide range of real-world situations. Nick Heiner, who leads the company’s work on these environments, stated that a single company’s Slack history can be more valuable than another firm’s combined emails, calendars, and chats if it captures conversations with sufficient variety and depth.
Real archives avoid common errors found in synthetic data, such as references to non-existent dates. However, the extent of performance gains from using real-world data remains unclear. No company has released tests comparing results from real-world data training against synthetic alternatives. This uncertainty leads to wildly different pricing, with buyers basing valuations on research theses that estimate the number of useful tasks contained in a specific volume of code.
Legal and Ethical Concerns
High-profile deals have already occurred in this space. In August, Google outbid the AI data company Mercor to purchase Spirit Airlines’ data as part of the carrier’s bankruptcy proceedings. The $10 million purchase included 100 million emails, 500 million Teams messages, and 176,000 employee records. Bloomberg reported in September that SpaceX’s AI unit is considering buying data from struggling startups.
These transactions face significant legal scrutiny. The flight attendants’ union at Spirit Airlines objected to the data sale on privacy grounds. Greg Ewing, an attorney who advises AI startups, has described such data as “legally radioactive.” While suppliers employ data scientists to scrub personally identifiable information and exclude human resources records, Ewing points out that private details such as doctor appointments and school pickup arrangements often remain in emails and calendars. He describes the current environment as a “Wild West of AI training” that is only beginning to transition toward proper licensing agreements.
The distribution of profits from these sales also raises ethical questions. Proceeds typically flow to senior creditors and investors rather than employees. Ifeoma Ajunwa, a law professor at Emory University and author of The Quantified Worker, describes this as a form of “captured capital.” She argues that employers should have express consent before selling workers’ data, noting that the legal justification for monitoring vanishes once a firm is out of business. Currently, there is no federal law banning such sales.
Founders have limited visibility into who buys their data and how it will be used. There is also little assurance that they are receiving fair offers. Data prices are expected to stabilize in the coming year, provided the AI market does not collapse. The firms that manage the end-of-life processes for startups will continue to appraise and sell the remaining digital assets.
Source: MIT Technology Review
