Until now, the biggest AI companies have been supported by constant injections of massive investment. But what happens when the funds stop coming in and they are expected to survive on their own income?
The fate of Character.AI, the controversial AI companion platform, could provide a warning.
On Tuesday, Character.AI filed for Chapter 11 bankruptcy, marking the beginning of an ignominious end for a once bustling startup valued at a billion dollars. Its bankruptcy filings, which were reviewed by 404 Mediasummarize critiques of the AI industry model. It gained a huge user base with a very attractive product, but racked up exorbitant compute bills without being able to monetize all the time people spent talking to its chatbots. All the while, it found itself mired in lawsuits alleging its products harmed its teenage users or even encouraged their suicides.
In the documentation, the company speaks frankly about its failures. CEO Colin Adams sums up the situation by saying that its “impressive technology is expensive to operate” and that it has consistently “struggled to achieve profitability and its operational expenses have been high.”
By the numbers, that’s $56.1 million spent on compute costs between January and August 2026, while spending an additional $19.7 million on employee salaries over the same period.
Its revenue remained dismal, dropping from $2.8 million in December 2025 to $5.3 million for August, while still generating an operating loss of $8 million to $9 million per month.
Desperate to stop the bleeding, the company resorted to “enshittification” of its product, as 404 puts it on. It began serving more ads, imposing stricter usage limits, and forcing cheaper “lobotomized” models on its users.
The company also “reduced its workforce,” Adams wrote. “In mid-2025, Character launched in-app advertising” and “in early 2026, Character also introduced in-app purchases.”
For a startup that once promised to bring “personalized superintelligence” to everyone on Earth, its business model and the experience of using its app were beginning to seem indistinguishable from a sleazy mobile game.
Attracting millions of users obviously does not guarantee long-term success. More users meant more computational costs, and the nature of how users interacted with chatbots – as companions – meant they were subjected to in-depth, hours-long conversations. It became a big bill to pay.
The same could be said for most major AI products. Many people have close relationships with ChatGPT or Claude. Power users leverage them to produce mountains of code, often running multiple instances at the same time to handle different tasks. In many cases, these AI providers have already changed their pricing systems to direct usage-based billing to address rising compute costs. OpenAI, meanwhile, began serving ads in ChatGPT to users on the free and low-paying tiers this year. And like Character.AI, it has also been the subject of several lawsuits alleging its chatbot drove users to suicide.
For now, leading AI companies can stay afloat by attracting huge investments (OpenAI and Anthropic) or subsidizing their AI operations with their traditional revenue streams (Google and Meta). But eventually there will come a time when the training wheels come off – and at that point, Character.AI’s demise might start to feel like a cautionary tale.
Learn more about AI: People are already forming romantic relationships with Meta’s muse
Gn bussni
