Entrepreneurs selling their large companies to BAT is the result of big tech monopolizing almost every inch of the consumer internet, and now a similar scene is happening in Silicon Valley's AI investment circle, with AI startups "selling themselves" to giants becoming a trend. On August 2nd local time, the well-known AI unicorn Character.AI announced a major "acquisition" agreement with Google, where Google will buy out the stock held by Character.AI's original investors at $88 per share, meaning the latter "acquired" Character.AI for $2.5 billion.
In fact, it's not surprising that Character.AI, a former AI unicorn, has reached the point of being sold. A month ago, overseas tech media The Information reported that the company was facing funding difficulties and seeking cooperation with Google, Meta, and Elon Musk's xAI to alleviate its predicament. Character.AI would share intellectual property rights with the cooperating companies in exchange for the right to use their computing resources.
In fact, Character.AI, an AI chatbot known for its role-playing capabilities, can be considered the most successful AI-native application at present. It has always been second only to ChatGPT in terms of traffic in the AI circle, with its core selling point being that users can "establish a virtual relationship" with AI.
With the rapid development of society and the intensification of contradictions, humans are feeling increasingly lonely. When needs and desires cannot be satisfied, seeking alternatives like AI and robots is the key to the popularity of AI chatbots today, and Character.AI is the one that has gone the furthest.
As an AI product known for user co-creation, Character.AI's UGC attributes are quite prominent. Even if users don't create their own virtual characters, the website provides a large number of characters created by other users, including celebrities from various industries such as Einstein, Musk, Zuckerberg, and Kanye West. AI companionship is currently the hottest AI product, possibly without exception, and Character.AI has always been a leader in this field, so the company naturally became an AI unicorn.
However, unfortunately, Character.AI can be said to be critically acclaimed but commercially unsuccessful. Since completing a $150 million financing round at a valuation of $1 billion in March 2023, Character.AI has not received even a dollar of funding. According to relevant institutional statistics, Character.AI's inference requests at its peak once reached one-fifth of Google's search traffic. Even in May this year, with many similar products appearing, Character.AI's visits were 277 million, an 18.89% month-on-month increase from April. In comparison, the AI search product Perplexity only had 30% of its visits, but Perplexity's valuation reached $3 billion.
In a sense, Character.AI is not lagging behind due to insufficient technical capabilities, ultimately leading to seeking a sale. As the CEO and co-founder of Character.AI, Noam Shazeer is one of the authors of the Transformer paper, which is the cornerstone of today's AI world, and the AI chat experience provided by Character.AI is top-tier. Unfortunately, the business prospects of AI companionship are bleak, with Character.AI having less than 100,000 subscription users and only $15.2 million in revenue for the entire year last year.
The core problem with AI companionship products is that the inference costs are extremely high, but the user's payment penetration rate is relatively very limited. Rashly increasing the payment level would lead to a decrease in users' willingness to pay, which has led to the failure of the pricing logic based on user scale that occurs in the internet world. At this point, the large user base is no longer a valuable asset for Character.AI, but rather a burden, so AI companionship may be more suitable for being "small but beautiful."
If Character.AI, as a "dream-making machine," fell into difficulties because it chose the wrong track, Google's "acquisition" this time almost perfectly illustrates what is called correct. The reason why "acquisition" is in quotation marks is that the transaction between Google and Character.AI is not the traditional acquisition model in the internet industry. Google did not take away Character.AI as a company this time, but obtained technology authorization from Character.AI, and the two founders of the company returned to Google and joined the DeepMind research team, while Character.AI's core development team of 30 people was incorporated into Google's Gemini AI department.
Character.AI said in the press release, "Most of Character.AI's excellent team will continue to stay on and will continue to build Character.AI's products to serve our growing user base." In fact, Google didn't want Character.AI's "shell" this time, and this operation is the most popular "reverse acquisition" or "employment-style acquisition" in the AI industry today.
After experiencing the severe internet anti-monopoly wave during the pandemic period, tech giants no longer choose to directly merge and acquire, but instead control products, technologies, and core personnel by investing in AI startups.
In fact, whether Character.AI found a "home" or Google got a bargain, this atypical acquisition case can only indicate that the AI bubble is starting to appear. Character.AI sought external cooperation and even sale because it had not been able to complete financing in the past year, resulting in a lack of funds to purchase computing power to maintain business development. The current situation is that high chip costs, gradually emerging data costs, and compliance requirements are beginning to put tremendous pressure on AI startups.
From Adept's CEO jumping ship to Amazon, Inflection's CEO leaving the company to join Microsoft, Character.AI's CEO and president turning to Google, to Microsoft executives saying that the big pie of AI investment needs 15 years to be rounded up, it seems that the AI bubble theory is becoming more and more real now.