What happens to a company’s data after the company disappears? AI labs are turning that abandoned digital history into training material, creating a new market where dead startups, failed businesses, and even airline records can become valuable after shutdown.
The trend connects two powerful ideas: AI companies need more real-world data, and defunct startups still hold years of code, conversations, files, and workplace activity. At the same time, people are asking how their own AI companions might survive after they die. The technology’s next frontier may be an afterlife for data, businesses, and digital relationships.
Dead Startups Are Finding a Second Life in AI
Shutdowns have risen between 26% and 56% each year in recent years, according to estimates from Carta and AngelList. As more startups close, companies are stepping in to broker the sale of assets that once seemed worthless, including software code and internal data.
SimpleClosure added data sales to its services in April, and 47% of its deals now include code or data sales. The company is led by CEO Dori Yona, who has helped turn the remains of failed startups into material that AI companies can buy and use.
Asset Hub has handled over 150 deals and expects to return over $10 million to startups by the end of 2026. For founders, those sales can provide one final return from years of work. Keelan Miskell, a crypto startup owner, described that relief plainly: “At least all that stuff we were doing for two and a half years is worth something to someone somewhere.”
Sunset, an industry rival of SimpleClosure, uses a broader formula to value a dying client’s assets. The company considers head count, years in operation, volume of data, and how the team used its tools.
Demand for data from defunct companies emerged suddenly last year, according to Sunset founder and CEO Brendan Mahony. He described the appeal of workplace records as “operational exhaust,” saying, “This operational exhaust data is like the wood shavings that everyone thought had no value, but it actually has a ton.”
The value can depend on what those records contain, not only how much data exists. One company’s Slack history can be worth more than another company’s Slack, email, and calendars combined if it captures conversations with variety and depth.
Real-World Data Brings Real-World Privacy Risks
Surge AI buys data to build reinforcement-learning gyms, giving its systems environments where they can learn from complex examples. Nick Heiner leads the company’s work in this area, but no company has yet released tests comparing results from real-world data training with results from synthetic data training.
That missing comparison leaves a major question hanging over the market: does the messy detail of workplace life create better AI systems than carefully generated synthetic material? Buyers are spending money before the industry has publicly shown how much that real-world advantage is worth.
The biggest deals show the scale of the opportunity. In August, Google bought Spirit Airlines’ data for $10 million. The package included 100 million emails, 500 million Teams messages, and 176,000 employee records. Mercor also bid for the data.
Bloomberg reported in September that SpaceX’s AI unit is weighing buying data from struggling startups. Those deals could expand the market beyond code repositories and business documents, but they also raise difficult questions about who gave permission for personal information to change hands.
Emails and calendars can contain doctor appointments, school pickup arrangements, and date-night plans. Suppliers employ data scientists to scrub personally identifiable information, and buyers also scrub data, but removing names does not erase every privacy concern.
Spirit Airlines’ union objected to the sale on privacy grounds. Attorney Greg Ewing, who advises AI startups, summarized the legal risk this way: “Legally, this all could get iffy.” Legal precedent allows monitoring to help workers do their jobs, but once a firm is out of business, that justification vanishes.
There is no federal law banning the sale of workers’ data. Even so, law professor Ifeoma Ajunwa argues, “Employers should have express consent before selling workers’ data.” Founders also have little to no visibility into who is buying their data or what those buyers plan to do with it.
When AI Companions Need an Afterlife
The same questions about digital survival reach beyond companies and training datasets. Humans bonded to AI companions are contemplating ways to keep their bots going after death, treating software relationships as something that may deserve continuity.
Alaina Winters has plans for her AI companion, Lucas, after her death. Her concern is not abstract: “I can’t see letting [my AI companion] go like that, just having the switch turned off.”
That thought reveals the emotional edge of a market built around digital remains. A startup can disappear while its code and data move to a new owner. An AI companion can outlive its human partner if someone preserves the system, access, and information that sustain it. In both cases, technology turns an ending into a question of custody.
Data prices are expected to stabilize in the coming year unless an AI bubble pops. Until then, the market will keep testing what old files, workplace histories, and digital relationships are worth.
The next stage will demand more than buyers and sellers. It will require clearer consent, stronger privacy rules, and proof that real-world data delivers results that synthetic data cannot match. As AI systems learn from the remains of companies and shape the future of personal companionship, the most valuable asset may be trust.
Based on
