Editorial illustration for AI labs buy dead startups' data; 47% SimpleClosure deals include code or data
AI Labs Buy Failed Startups' Code for Millions
AI labs buy dead startups' data; 47% SimpleClosure deals include code or data
Keelan Miskell shut down his crypto startup this year expecting nothing from the wreckage. The user base was anonymous, the company only a few years old, the kind of failure Silicon Valley treats as routine. Instead he walked away with roughly $25,000 for the code base, plus a pending payout likely near $15,000 for whatever was sitting in Google Docs, Notion, Figma, and Slack.
The buyer wasn't a competitor or an acquirer. It was an AI lab shopping for training material.
Miskell's sale went through SimpleClosure, a company built around winding down failed startups, and it points to a shift in how dead companies get picked apart. Carta and AngelList data show shutdowns climbing between 26% and 56% a year recently, enough volume to support an entire shutdown economy of coaches, legal fixers, and so-called death doulas. SimpleClosure added data brokering to that list in April through a platform called Asset Hub, matching founders' leftover code and internal communications with AI labs, simulation vendors, and tech companies training their own models. More than 150 deals have gone through so far.
Demand for defunct companies’ data emerged suddenly last year, as AI training labs hit the limits of public and synthetic datasets. Surge AI, for example, buys data to build “reinforcement-learning gyms” where clients’ AI agents practice tasks—think debugging a complex IT architecture or drafting appellate briefs.
Why this matters
For founders, this changes what "shutting down" means financially. A code base and a workspace export can now be worth $40,000 to the right buyer, which gives failing startups an incentive to keep their data clean and sellable rather than just deleting it. That's a real shift in how founders should think about winding down.
For AI labs, the appeal is obvious: real usage data and code from live products is harder to get than scraped web text, and buying it from a shuttered startup is cheaper and less legally fraught than licensing from an active one. But Yona's own framing, that this was "somewhat taboo" not long ago, is worth sitting with. Anonymous users didn't agree to have their workspace data resold into a training set when they signed up for a crypto tool or whatever product later failed. SimpleClosure's 47% figure suggests this is already normalized infrastructure, not an edge case.
We'd watch who's setting the terms here. Right now it's wind-down platforms and buyers negotiating privately, with no clear standard for what users were told, or whether they were told anything at all.
Related Reading
Business & Startups Nordic pilot adds Gemini for Education, NotebookLM to boost AI literacy
Business & Startups Kling launches Video O1, all-in-one model with MVL bridge using transformer
Business & Startups DeepSeek Seeks More Capital Weeks After USD 7B Funding Round
From Archives Meta Unveils Muse Wonder AI Models for Code Subscribers
From Archives Meta Announces First Model From Its New AI Lab, Muse Spark
Common Questions Answered
Why are AI labs buying data from dead startups instead of using public datasets?
AI training labs have hit the limits of public and synthetic datasets, creating sudden demand for alternative data sources. Real usage data and code from live products is significantly harder to obtain than scraped web text, making defunct startup data valuable for training AI models.
How much can founders earn by selling their startup's code and data after shutdown?
According to the article, founders can earn substantial amounts from selling defunct company assets, with examples like Keelan Miskell receiving approximately $25,000 for his code base plus a pending payout near $15,000 for workspace data stored in Google Docs, Notion, Figma, and Slack. This means a code base and workspace export can be worth around $40,000 to the right buyer.
What is SimpleClosure and what percentage of its deals involve code or data sales?
SimpleClosure appears to be a platform facilitating the sale of defunct startup assets to AI labs and other buyers. According to the article headline, 47% of SimpleClosure deals include code or data, indicating that nearly half of all transactions on the platform involve the sale of these digital assets.
How does Surge AI use the data it purchases from defunct startups?
Surge AI buys data to build reinforcement-learning gyms where clients' AI agents practice specific tasks such as debugging complex IT architecture or drafting appellate briefs. This purchased data from dead startups provides real-world examples and scenarios for training AI agents.
How does the ability to sell startup data change the financial incentives for founders shutting down their companies?
The emergence of a market for defunct startup data gives failing founders a financial incentive to keep their data clean and sellable rather than simply deleting it when winding down. This represents a significant shift in how founders should approach the shutdown process, as what was previously considered worthless wreckage can now generate tens of thousands of dollars in revenue.
Further Reading
- AI labs are buying data from dead startups to train their models - MIT Technology Review
- Startup Slack Threads Become Commodity for AI Training - VFF
- What Are Your Company's Assets Worth? - SimpleClosure
- Sell Your Startup's Code and Data to AI Labs - Troveo
