A market most acquirers ignore
Most corporate development programs look for healthy companies with willing owners. Wind-downs are another source: products that worked but couldn't raise their next round, customer bases that need a new home, and teams that built something useful. These assets often sell as asset purchases, with no need to take on the company's liabilities.
In October 2026 MIT Technology Review reported that AI labs and data vendors are buying code and workspace data from shut-down startups, through wind-down platforms such as SimpleClosure and Sunset. That turns wind-downs into a contested market, and raises what founders and investors expect to receive.
What can be bought
| Asset | Why buy it | What to check |
|---|---|---|
| Code and product | Skip a build, or add a feature | Ownership, open-source licenses, contractor IP assignments |
| Customers and contracts | Revenue and relationships | Assignment clauses and customer consent |
| Domain, brand and content | Traffic and search history | Trademark status and any liens |
| Data | Training, analytics or product value | Privacy terms, consent and contracts |
| Team | An acqui-hire of people who know the product | Offers made directly to people, not 'bought' |
Moving quickly without getting hurt
- Find out who controls the sale: the board, the investors, secured lenders or an assignee.
- Buy assets rather than shares to leave liabilities behind.
- Get legal review of IP ownership and data rights before money moves.
- Plan customer communication on day one, so revenue doesn't leave with the old brand.
On the selling side, we explain how founders can sell a startup's data and code before shutting down, and our news analysis of AI labs buying from dead startups covers the trend.
Frequently asked questions
Is buying from a shutting-down startup risky?
It can be if ownership of the IP or data is unclear. Asset purchases with legal review of IP assignments, licenses and data consents manage most of the risk.
Sources
- AI labs are buying data from dead startups to train their models · MIT Technology Review, Oct 13, 2026