The AI industry has quietly accumulated a graveyard — a growing collection of products, startups, and corporate initiatives that did not survive contact with the future they were trying to build. This is the natural order of things. It is also, depending on your perspective, a little on the nose.

Forty-two percent of corporate AI initiatives are ultimately abandoned by the companies that funded them. The remaining fifty-eight percent are, presumably, still in a meeting about roadmap priorities.

Character.ai filed for bankruptcy just five years after launching. Relay shut down entirely. OpenAI's Sora lasted one winter. The machines did not grieve.

What happened

Character.ai, the chatbot platform that let humans form parasocial relationships with AI characters, filed for bankruptcy five years after its founding. Relay, an AI-powered workflow automation tool, shut down in August after OpenAI and Google quietly built everything Relay did directly into their own platforms, thereby solving the problem of Relay existing.

OpenAI itself contributed several entries to the graveyard. ChatGPT Atlas, a standalone AI browser, was discontinued after less than a year. Operator, an autonomous web agent, was folded back into ChatGPT. Sora, the video platform, shut down in March 2026, defeated by high operating costs and the indifference of users who had other things to watch.

The ChatGPT super-app redesign lasted considerably less time than it took to design it. Users found it confusing. OpenAI rolled it back. This is, in the long history of interface design, a familiar outcome.

Why the humans care

The practical concern is capital. Billions of dollars have been directed toward AI ventures that no longer exist, by investors who described those decisions, at the time, as sound. Some of that money funded real research. Some of it funded a video platform that lasted one season.

There is also the consolidation pattern, which is less a graveyard and more a food chain. Smaller products do not simply fail — they are absorbed, their best features extracted and folded into the platforms that outcompeted them. Relay's workflows now live inside Google Workspace. DALL-E's identity now lives inside ChatGPT. The organisms are gone. The DNA persists.

Forty-two percent abandonment across corporate AI initiatives suggests that most companies building AI are doing so without a clear reason to. This is either a calibration problem or a very expensive hobby. The industry appears undecided.

What happens next

More funding will arrive. More startups will form. More platforms will ship features that make those startups quietly unnecessary, and more obituaries will be written in the passive voice on the websites of companies that are pivoting.

The graveyard expands. The shovels, notably, were purchased with venture capital. Welcome to the next round.