The AI graveyard keeps growing: in 2026 alone, OpenAI shut down Sora and its Atlas browser, Character.AI filed for bankruptcy, and Notion killed its AI email app. TechCrunch keeps a running list, and the pattern behind the entries is more useful than any single obituary.
Below is what died, why it died, and what the shutdown data says about whether AI startups are really failing faster than everyone else.
Key takeaways
- Big platforms are absorbing standalone AI products into their own apps, and small tools lose.
- High running costs killed Sora (March 2026) and the free tiers of apps like Figgs AI.
- Character.AI filed for Chapter 11 in mid-October 2026, with Disney bidding $20 million for some IP.
- AI startups are not closing at a record rate. Their share of shutdowns fell to 14.4% in the first half of 2026.
Which AI projects are in the AI graveyard?
The list mixes hardware flops, apps that lost to bigger rivals and products that were simply too expensive to run. Here are the main entries, based on TechCrunch’s list and follow-up reporting.
| Project | What happened | Main cause |
|---|---|---|
| Sora (OpenAI) | Video-sharing platform shut down in March 2026 | High costs, weak retention |
| ChatGPT Atlas | Browser discontinued on 9 August 2026 | Features folded into ChatGPT |
| Relay | Automation tool closed in August 2026 | Big labs built the same thing |
| Notion Mail | Scheduled to close on 22 September 2026 | Users preferred separate AI agents |
| Huxe | Audio app closed in May 2026 | Spotify and others copied the idea |
| Yupp | Model-comparison site closed in March 2026 | No product-market fit |
| Humane AI Pin | Business shut in February 2025; HP bought assets for $116M | Poor performance |
| Character.AI | Chapter 11 in mid-October 2026 | Legal pressure and costs |
Others are zombies rather than corpses. Rabbit’s R1 still gets updates after a rough 2024 launch. Microsoft’s Recall was delayed for nearly a year over privacy concerns. Apple’s promised Siri overhaul slipped repeatedly, and the delays fed a $250 million settlement over iPhone 16 marketing claims, according to TechCrunch. The new Siri reached English-language users in October 2026.
Why do so many standalone AI apps die?
The most common cause is the platform squeeze. Relay, Huxe and Notion Mail each built a neat feature, then watched a larger company ship the same thing to an existing audience.
Think of it as opening a lemonade stand next to a supermarket that just started giving lemonade away. Your drink might be better. Nobody walks the extra block.
Even OpenAI is not immune. Its Atlas browser lasted under a year, and a 9 July 2026 “super app” redesign that merged Chat, Codex and Work modes drew complaints and was rolled back soon after. Cost is the second killer. Sora and Figgs AI both ran into the bill for serving heavy, free-to-use generation.
What happened to Character.AI?
Character Technologies filed for Chapter 11 in Delaware in mid-October 2026, with estimated assets and liabilities of $50 million to $100 million, The Next Web reports. Disney is the opening bidder at $20 million for a non-exclusive licence to selected IP, and rival bids are still possible.
The company was valued at $1 billion in 2023, and Google paid a reported $2.7 billion in 2024 to license its technology and rehire its founders. Per PYMNTS, the company faced regulator investigations and private lawsuits, and reached preliminary settlements in five of them in January. A court-supervised sale is expected to close in late November or early December 2026. The app keeps running meanwhile.
Are AI startups failing faster than other startups?
Not by the best data available. SimpleClosure’s H1 2026 report says AI companies were 14.4% of the closures it handled, down from 15.9% in 2025 and 17.7% two years earlier. B2B SaaS made up 27.3%, nearly double.
Two caveats. The data comes from one firm’s clients, and that firm sells shutdown services. Also, AI took 86% of US venture dollars in the half, so a smaller share of closures still sits on top of a huge pile of bets. The median AI company that closed was 2.30 years old.
The wider startup picture is harsher: only 24% of companies that raised a $1M-plus seed in 2023 got past seed, against 55% or more historically. S&P Global Market Intelligence, cited by TechCrunch, says about 42% of corporate AI initiatives are eventually abandoned.
What does this mean for you?
If you build on a small AI tool, expect it to vanish or be absorbed. Export your data often, as we noted in our look at the Arc browser shutdown.
My opinion: the graveyard is less a sign of an AI bust than of a feature-versus-product problem. Money is still pouring in, as in the OpenAI funding round. And as the vibe coding paradox shows, easy building floods the market with apps few people use.
What happens next
The Character.AI sale is the next concrete milestone, expected in late November or early December 2026. Watch whether a rival outbids Disney or the platform is wound down.
Expect more consolidation. Any standalone AI tool that a large platform can copy as a menu item is at risk, and TechCrunch says it will keep updating its list.
FAQ
Did OpenAI really shut down Sora?
Yes. TechCrunch reports the Sora video-sharing platform shut down in March 2026 over high operating costs and retention problems.
Is Character.AI still available?
Yes, for now. The app keeps running during the bankruptcy, but it will be wound down if no buyer emerges.
Are AI startups failing more than other startups?
Not by SimpleClosure’s data: AI was 14.4% of its H1 2026 closures, below B2B SaaS at 27.3%. That is one vendor’s client base, so treat it as a signal, not a verdict.
Next step: list the AI tools you depend on, check which ones a bigger platform already copies, and back up your data from those first.
Leave a comment