Character AI has just been acqui-hired by Google, similar to how Inflection AI was recently acqui-hired by Microsoft.
Character AI is an innovative generative artificial intelligence platform that allows users to create and interact with custom AI characters. Launched in September 2022 and headquartered in Menlo Park, California, Character AI was developed by Noam Shazeer and Daniel De Freitas, former developers of Google's LaMDA. The platform has garnered significant popularity, especially among younger demographics, with a user base exceeding 20 million by early 2024.
Character AI has a notable engagement rate, with users spending an average of 120 minutes per day on the platform, the highest among similar AI chatbots. Additionally, the platform boasts a high retention rate compared to typical mobile apps, which generally see a 30-day retention rate of around 3-4%. This sustained user engagement is attributed to the platform's ability to offer personalized experiences through its AI characters.
The platform's monthly visits reached approximately 208 million as of April 2024, highlighting its rapid growth and widespread appeal. It has also been downloaded over 10 million times globally from app stores, indicating strong user interest and satisfaction. Character AI's revenue is projected to be around $16.7 million annually in 2024, with a valuation reaching up to $5 billion.
Google's acquisition of Character AI is part of the tech giant's broader strategy to expand its generative AI capabilities. The acquisition will provide Google with access to Character AI's cutting-edge technology and talent, enhancing its existing AI offerings and strengthening its competitive position in the market.
Why It Matters
The generative AI hype is receding. Many startups have been struggling to find a sustainable business model, as training models is costly and revenues are still not on par with capital expenditures. Heavy generative AI players need GPUs to train and serve models, and that is costly. As burn rates increase very fast, the only solution that often remains is acquisition by a big cloud provider. The benefit for the startup is access to a wider pool of GPUs, in-house expertise, and existing distribution networks. The benefit for the cloud provider/big tech is intellectual property, mostly, and an innovative product that can complement existing offerings and probably could not have been created due to the bureaucratic effects of large organizations.
What It Tells About Generative AI
Generative AI is useful but often difficult to monetize. Freemiums, like the one offered by ChatGPT, are great but only work when you have a very large pool of users willing to pay for access to the most advanced models. Even then, it might not be enough to recoup the cost of training and serving models. OpenAI is the perfect illustration for this. Because of the hype, a lot of capital expenditure has been consumed, and there aren't enough use cases or adoption to recoup the investments. Some use cases are even flawed because they are built on incorrect assumptions about what generative AI models can achieve. Additionally, we now have state-of-the-art open-source generative AI models (Llama 3 & co), making it even more difficult for some startups to make money off serving models since any company can deploy their own model for their specific internal use case. Generative models are a commodity now, and I recently wrote about that on my Lycee AI blog.
Generative AI models are useful but only when used appropriately. In the coming months, we will see more generative AI startups being acquired and many going bankrupt as more people realize how unproductive generative AI solutions can be when they are poorly designed. We are reaching the trough of disillusionment. Only after that will we reach the plateau of productivity.
