AI Labs Pay Millions for Data From Dying Startups

Keelan Miskell, a former crypto‑startup founder, sold his company’s code base and workspace data through SimpleClosure for roughly $25,000, plus a pending $15,000, proving that even anonymous user archives can fetch cash.

Key points

• SimpleClosure’s Asset Hub handled 150+ deals, expecting $10 million to startups by 2026.
• Sunset, a rival, values data by head count, years, volume, and tool usage.
• Surge AI buys “reinforcement‑learning gyms” from real archives, citing varied Slack histories.
• Google outbid Mercor for Spirit Airlines’ 100 million emails and 500 million Teams messages, paying $10 million.
• Legal experts call private‑data sales “legally radioactive” and warn of privacy risks.

What happened

The trend began when SimpleClosure added data sales to its services in April. Its Asset Hub platform lets founders sell code, emails, chats, and documents to buyers such as foundation AI labs, vendors building simulated workspaces, and mature tech companies. CEO Dori Yona notes that “once rare and ‘somewhat taboo’ become increasingly routine, with 47% of SimpleClosure deals now including code or data sales.”

Sunset, another company in the niche, follows a similar model. Founder Brendan Mahony explains that a dying client’s assets are valued by head count, years in operation, data volume, and how the team used its tools. He likens “operational exhaust data” to wood shavings, saying, “This operational exhaust data is like the wood shavings that everyone thought had no value.”

Demand surged last year as AI labs hit limits of public and synthetic datasets. Surge AI, for example, purchases real‑world archives to train agents in complex tasks such as debugging IT architecture or drafting appellate briefs. Nick Heiner, who leads the company’s environment work, says he seeks data reflecting a range of situations, from disagreements to tricky judgment calls. He notes that one Slack history can be worth more than another’s Slack, email, and calendars combined.

Large deals have followed. In August, Google outbid Mercor for Spirit Airlines’ decades‑old data during its bankruptcy, paying $10 million for 100 million emails, 500 million Teams messages, and 176,000 employee records. Bloomberg reported in September that SpaceX’s AI unit is weighing buying data from struggling startups.

The backstory

Shutdown rates have risen between 26% and 56% annually, according to Carta and AngelList. An end‑of‑life ecosystem—comprising coaches, “death doulas,” and firms handling legal, tax, and operational matters—has emerged to help companies wind down. Data resale is a natural extension of this ecosystem, turning otherwise discarded digital footprints into revenue streams.

Drivers of the market

The primary driver is AI training’s appetite for high‑quality, diverse data. Real archives avoid synthetic errors like impossible dates, and they provide context that synthetic data cannot replicate. However, the performance gains remain unproven; no company has released comparative tests. Pricing reflects this uncertainty, with buyers quoting wildly different numbers based on “research theses” that estimate how many useful tasks arise from, say, 100,000 lines of code.

Legal and ethical implications

Greg Ewing, an attorney advising AI startups, calls such data “legally radioactive.” Suppliers scrub personally identifiable information, but buyers claim they do so as well. Still, private details—doctor appointments, school pickups, date‑night plans—can lurk in emails and calendars. The Wild West of AI training is only beginning to yield proper licensing agreements and vetted use, raising questions about who ultimately profits. Proceeds typically flow to senior creditors and investors, not employees, a form of “captured capital” noted by law professor Ifeoma Ajunwa.

What builders and businesses should do

• Map out all code and workspace data before winding down.
• Engage a data‑resale partner early to negotiate fair valuation.
• Ensure robust PII scrubbing and compliance with privacy laws.
• Draft clear licensing terms for any future AI buyers.
• Consider whether retaining a data‑asset portfolio could provide post‑exit revenue.

The bottom line

Watch for regulatory clarifications on data resale and AI training use. As AI labs continue to pay top dollar for real archives, the market for defunct company data will mature, reshaping exit strategies for founders and investors alike.