TechCrunch has begun keeping a running list of AI projects that shut down or failed to meet expectations. The name is direct: the AI graveyard.
Lists like this appear when a sector matures. Shutdown stories being compiled alongside funding news shows the field is no longer told purely as a growth story.
Compiling shutdown stories is itself a signal. Once failure stories start being written in a field, that field has left behind the period when every idea got funded.
Who is on it
The two standout examples represent two different kinds of failure in the field.
The first is Apple's repeatedly delayed Siri overhaul. The project did not shut down, it kept slipping; the outcome was the same, because competitors shipped their products in that time.
The second is OpenAI's messy super app launch. That is a different pattern: the product shipped but did not meet expectations.
| Form of failure | What happens |
|---|---|
| Continuous delay | The product never ships, the window closes |
| Missing expectations | The product ships, the response does not come |
| Shutdown | The project or company ends entirely |
The third pattern forms the core of the list: startups that shut down. These are usually companies built on a single product idea that a model update rendered unnecessary.
Why such a list now
The number of companies founded in AI over the past two years is higher than in any comparable period. The attrition following a founding wave of that size arrives at the same scale.
The list functions as a kind of memory. Closed projects are forgotten quickly and the same ideas get funded again a few years later; keeping a record at least makes that cycle visible.
It serves another function: calibrating expectations on the investor side. When assessing the third startup arriving with the same idea, remembering why the earlier ones did not work changes the quality of the decision.
What it cannot measure
A list like this has a built-in gap: it collects only what is visible. Small startups that close without press coverage do not appear, and they are far more numerous.
The second gap is timing. How long to wait before calling a project a failure is undefined; a delayed product can ship two years later and work.
A third gap concerns causes. The list says what closed but not why it closed, and the why is the instructive part.
The lesson in it
What the two examples share is that neither belongs to a company short of resources. The cause of failure is not money but timing, focus and expectation management.
For a small team the practical conclusion is this: delayed projects at large companies show that windows in this field do not close as fast as assumed. Where Apple lost ground by postponing, far smaller teams that shipped found room.