Updated October 2026
Selling a closed company's codebase
A closed company's codebase used to interest only a buyer who wanted to run the product. AI buyers now want repositories as training material, whether or not anyone will ever deploy the code again. MIT Technology Review reports that one shut-down crypto startup sold its code base alone for about $25,000.1
How much is out there
A market measured in repositories and commits
SimpleClosure says companies on its AssetHub have offered nearly 500M lines of code across 3,000 repositories, with 1.7M commits, 200,000 pull requests and 3,200 contributors. Those are SimpleClosure's own figures, and they describe what has been offered, not what has sold.2
The crypto startup in MIT Technology Review's report had run for about two and a half years. Its code base found a buyer anyway, which suggests, in our reading, that a long history is helpful but not required for code.3,1
| Fact | Value and source |
|---|---|
| A closed crypto startup's data | About $25,000 for its code base, with about $15,000 for workspace data pending |
| Code offered on AssetHub | Nearly 500M lines of code across 3,000 repositories, with 1.7M commits, 200,000 pull requests and 3,200 contributors |
| Code that drops out | Roughly 37% of raw code disappeared after removing third-party and generated code and internal duplication |
| What sets the value | Value depends on scale, quality, differentiation and scarcity; more employees and more years operating raise it, and testing and documentation can matter more than line count |
What counts
Not every line in the repository counts
SimpleClosure says roughly 37% of the raw code it looked at disappeared once third-party code, generated code and internal duplication were stripped out. Vendored libraries, build output and copied modules are not the company's own work, and buyers remove them before they value anything.4
A quick test before talking to buyers: run a line counter that skips vendored, generated and build directories, and compare its total with the raw count. If the gap is near SimpleClosure's 37%, the repository is ordinary. If it is much wider, expect a buyer's working figure to fall well below the number on the dashboard.4
SimpleClosure also says testing and documentation can matter more than line count, and that more employees and more years in operation raise value. Our reading: a modest repository with tests, a readable commit history and real review threads may compare well with a much larger one padded with generated files.5
Who buys
Two channels that take code from closing companies
Turing says its Project Lazarus licenses code, tickets, communications and decision records from operating and winding-down companies, and pays up to $1M in one payment. That ceiling is Turing's own claim; Turing resells to AI labs.6
SimpleClosure says its AssetHub buyers are AI foundation labs, reinforcement learning environment builders and other technology companies. Our reading is that environment builders are one reason code with its history attached draws interest: the issues, pull requests and review comments show how engineers reasoned about a change.7
Preparation
Before anyone sees a repository
Search the code and its full history for secrets: API keys, passwords, tokens and connection strings. Rotate or revoke anything that still works. List the open-source dependencies and their licenses. Note which repositories hold work done for a particular customer, or customer data in test fixtures.
SimpleClosure says sensitive information and credentials are de-identified in every dataset before it is licensed. Ask any buyer how it does the same, who runs the tool, and whether the company sees the cleaned output before it leaves. Then decide whether to offer the full history or a snapshot; keep the history if you can.8
FAQ
Questions, answered with sources
SimpleClosure says roughly 37% of raw code disappeared after removing third-party and generated code and internal duplication.4
SimpleClosure points to scale, quality, differentiation and scarcity. More employees and more years operating raise value, and testing and documentation can matter more than line count.5
SimpleClosure says nearly 500M lines across 3,000 repositories, with 1.7M commits, 200,000 pull requests and 3,200 contributors.2
We do not answer legal questions. Ask the licensing company: their team reviews this with you before anything is signed.
Before you sign
Legal questions go to the licensing company
We do not answer legal questions. Ask the licensing company before you sign: their team reviews these with you.
Questions for this kind of business
- Do the open-source licenses in our dependencies affect licensing the repository as training data?
- Did any customer agreement give that customer rights in code we wrote for it?
- Were every contributor's rights in the code assigned to the company?
- Who owns this data: the company, an estate, or someone else?
- Do customer, vendor or platform contracts, NDAs or API terms restrict this?
Questions for any buyer
- Is the license exclusive, and for how long?
- Can you sublicense the data, and to whom?
- Which AI labs will receive it, and will you tell us?
- How long do you keep it, and how is it deleted?
- What is your liability cap, and who indemnifies whom?
- Will the contract ban re-identification of people in the data?
- Who removes personal data, by what method, and do we approve it first?