The AI graveyard is the unpaid bill of a growth-first model

The AI graveyard is the unpaid bill of a growth-first model

Character.ai filed for bankruptcy in October, five years after it launched. Ollie, an AI assistant for families, shut down on November 6, about five months after its debut. Both now sit on TechCrunch’s running list of AI casualties, alongside stumbles from OpenAI, Apple, and Microsoft.

A list like that invites the conclusion that the technology is overrated. Its entries say something narrower. In the most prominent cases the AI worked, and the product ran out of a reason to exist as a business: it cost too much to run, could not keep the users it won, or could not raise the next round. That is the bill for growing first and working out the economics later.

What the Entries Have in Common

The scale comes first. S&P Global Market Intelligence estimates that about 42 percent of AI initiatives are ultimately abandoned by the companies that started them. TechCrunch lists the reasons as insufficient funding, technical challenges, competition, difficulty scaling, and weak user demand. Only one of the five is squarely about the technology.

The individual cases follow that proportion. OpenAI’s video platform Sora shut down in March after struggling with high operating costs and with user retention. Ollie’s chief executive cited an inability to stand out, a failure to grow, and not raising enough capital. Relay, a five-year-old workflow automation tool, closed in August once OpenAI, Google, and other large platforms had built similar automation into their own products.

Character.ai is the largest example. It raised $150 million at a $1 billion valuation in 2023, and a year later Google paid a reported $2.7 billion to license its technology and bring its founders back. The app is still running during the bankruptcy, which the company attributed to legal, regulatory, and operational headwinds. An audience turned out to be a different asset from a business.

The Order of Operations

A familiar sequence lies behind those endings. An early-stage subscription company spends most of its time and money on one problem, acquiring customers faster, in order to prove that it has a viable business. As soon as that problem is solved, someone in finance notices that the total customer count and the committed monthly recurring revenue are declining1.

AI products meet that sequence carrying an extra weight, because every use costs money to serve. A product that is expensive to run needs users who stay. Sora, by TechCrunch’s account, had the expense without enough of the staying.

Not every entry fits. Some are features that shipped late or met resistance, such as Apple’s delayed Siri upgrade and Microsoft’s Recall, which drew an immediate privacy backlash. And OpenAI’s retreats were often consolidations. Its Atlas browser lasted less than a year before its most useful features were absorbed into ChatGPT, and its Operator agent was folded in the same way. A large company can fold a failed product into a successful one, and a startup with one product cannot.

What Gets Funded Next

If the same arithmetic keeps applying, the list will probably keep growing in a particular shape: stand-alone products built on a feature that a large platform can bundle, and consumer apps whose costs rise with every user. Investors may well ask for retention and cost-to-serve figures at the stage where they once asked only for signups. Founders could find that the fundable pitch is a product whose users pay more than they cost within the first year.

TechCrunch calls its list a record of what these attempts can teach the rest of the industry. Growth can postpone the question of whether a product pays for itself, but it has never answered it.

Sources

1 Mehta, Nick, Dan Steinman, Lincoln Murphy. Customer Success.

“The AI graveyard: a running list of projects and startups that didn’t make it.” techcrunch.com.

Leave a Reply

Your email address will not be published. Required fields are marked *