Since 2023, hundreds of AI startups have shut down, sold at a loss, or had their founding teams poached by the big players. Humane AI alone raised $230 million and sold to HP for $116 million, according to The Verge. That is a $114 million hole in investor pockets. And Humane is just one name on a very long list.
The Scale of the Problem
The AI funding wave that started in 2022 was historic. According to CB Insights, global AI startup funding topped $50 billion in 2023 alone. Valuations stretched beyond reason. Series A rounds that would have taken 18 months of proven traction got done in six weeks on a demo day deck.
The hangover hit fast. By 2025, according to PitchBook data, AI startup failures were running at a pace not seen since the dot com collapse. Founders who raised at 100x revenue multiples found themselves unable to raise follow on funding at any multiple. Quietly, the lights went out.
The companies that survived were not the ones with the best pitch decks. They were the ones with paying customers, defensible data, or both.
A Tour Through the Graveyard
The Wrapper Graveyard
This is the largest category by far. These startups built products on top of GPT-4 or Claude and charged monthly subscriptions. The problem is that OpenAI, Anthropic, and Google are not just API providers. They are competitors who study what you built and ship it natively six months later. Jasper AI was valued at $1.5 billion in 2022, according to TechCrunch. By 2025, the company was a shadow of that size after ChatGPT absorbed its core writing assistant use case.
The lesson is not subtle. If your entire value is “ChatGPT but with a better UI,” you do not have a business. You have a feature that lasts until OpenAI decides to ship it for free.
The Hardware Graveyard
Humane AI built a wearable AI pin. The concept was real. The execution flopped. Reviewers called it slow, unreliable, and unnecessary. The company raised $230 million, according to Crunchbase, and sold to HP for $116 million. The Rabbit R1 followed a similar arc: massive preorders, brutal reviews, and a product that never delivered on its promise.
Consumer AI hardware is harder than it looks. The margin for error is near zero because people return something that does not work in the first week. Software companies get the benefit of iteration. Hardware companies get chargebacks.
The Foundation Model Graveyard
Building a frontier model costs hundreds of millions of dollars. Only a handful of companies in the world can sustain that spend. The rest got absorbed. Inflection AI, one of the most hyped model startups with backing from Bill Gates and Reid Hoffman, lost both of its co-founders to Microsoft in 2024, according to Bloomberg. Adept AI’s team was effectively absorbed by Amazon the same year.
The value was not in the company. It was in the people.
This is the graveyard you do not hear about as often because technically the company still exists on paper. But when the founding team is gone and what remains is a shell, the startup is dead. The acquisition just has better PR than a shutdown notice.
What This Means for You
Here is what I think when I look at this graveyard. The average person reads these failure stories and thinks “AI is overhyped.” The sharp operator reads the same stories and asks: “What survived and why?”
The companies that made it through had one thing in common. They built something that would be painful to rip out. Vertical AI tools embedded in daily workflows. Companies with proprietary training data nobody else could replicate. Platforms that got smarter with user data in ways the generic models could not match.
If you are building anything with AI today, that is the only question worth asking. What happens if OpenAI ships this feature for free tomorrow? If the answer is “my business is over,” you are not building a company. You are building a feature.
If you are choosing AI tools rather than building them, the graveyard teaches a different lesson. Buy tools with proven retention, not tools with impressive demos. Check whether the company has been around for at least two years. Read reviews from people who actually pay monthly. Platforms like AppSumo surface vetted tools with lifetime deal pricing, which filters out a lot of the noise since a tool needs a real user base to get listed there at all.
For video creators and content businesses, this shakeout matters. Many AI video startups that raised big in 2023 never made it to 2026. The ones still standing, like InVideo AI, built genuine product depth rather than riding early hype. When you pick a tool to anchor your workflow around, you want a company that has already survived a few cycles, not one hoping to survive the next one.
The Bottom Line
The AI graveyard will keep growing. Not because AI is a fraud, but because most of the people who raised money to build AI companies were not building businesses. They were building features and calling them companies. The correction was not just predictable. It was necessary. The money flowing into real survivors now is better quality capital chasing better quality fundamentals. That is how a market matures. The question is whether you are learning from the graves or waiting to dig one.
Frequently Asked Questions
What is the AI graveyard?
The AI graveyard is an informal term for the growing list of AI startups that have shut down, been acquired at a loss, or had their founding teams absorbed by larger companies since 2023. It includes consumer hardware companies, AI writing tools, and foundation model startups.
Why did so many AI startups fail?
Most failed because they built thin products on top of APIs from OpenAI or other providers. When those providers shipped the same features natively, the startups lost their reason to exist. Others failed because they raised at valuations that required explosive growth that never came.
Which major AI startups shut down or were absorbed?
Humane AI sold to HP for $116 million after raising $230 million, according to The Verge. Inflection AI lost its founding team to Microsoft. Adept AI’s team was absorbed by Amazon. These are the high profile names. Hundreds of smaller AI startups closed with no announcement at all.
Is AI still worth investing in after so many failures?
The failures weeded out the weakest companies. The ones that remain have stronger fundamentals and more defensible positions. The question is not whether to pay attention to AI. The question is which companies built something that would be painful to stop using.
How can I find AI tools that will actually last?
Look for tools that have been around for at least two years and have real paying users. Check whether the core feature can be replicated by ChatGPT or another generic model overnight. Tools with deep workflow integration or proprietary data tend to outlast tools that are purely interface wrappers around someone else’s model.
