On December 31st, according to The Information, citing two sources familiar with the matter, the company's annualized revenue (annual recurring revenue) recently exceeded $1.6 billion, benefiting from the strong growth in ChatGPT users, a significant increase from CEO Sam Altman's disclosure of $1.3 billion in mid October.
Due to the high cost of training large language models, industry leader OpenAI is also seriously exploring its feasible business model. According to the official website of OpenAI, its products are currently mainly divided into two categories: one is products mainly based on application programming interface APIs, including callable GPT models and DALL · E models (cultural and graphic models); Whisper (speech recognition model) and a series of features such as fine-tuning provided for developers; Another type is products centered around ChatGPT dialogue robots, with the personal version being extremely popular among the public, while the enterprise version has not yet announced its price.
Since the beginning of this year, OpenAI has gradually demonstrated its commercial prototype from its products. At present, it earns its main income monthly through ChatGPT Plus's membership subscription service and allowing software developers to access its models through APIs.
Among them, the membership version subscription service of ChatGPT Plus was launched by OpenAI at the beginning of the year, and users only need to pay $20 per month to enjoy faster response speed, prioritizing the use of GPT-4 models and many other features. According to the latest industry data, the number of paying users for ChatGPT Plus is estimated to be between 230000 and 250000.
It should be pointed out that the API based billing model is currently the most relied upon business model for OpenAI. In specific implementation, OpenAI charges based on input and output traffic (or tokens, word elements), and its official website has shown some models of traffic billing solutions.
Image source: OpenAI official website
Chen Shi, investment partner of Fengrui Capital, told Interface News that according to the median level of GPT-3.5 Turbo traffic when using AI applications, as long as a user uses the application (DAU) every day, the app company behind the user needs to pay OpenAI about 0.2 yuan in traffic fees. By the same token, if a daily active APP application with millions of users connects to the GPT port, a daily traffic fee of 2 million RMB must be paid to OpenAI.
In addition, after collaborating with Microsoft, Microsoft's cloud service Azure OpenAI can directly call the OpenAI model, including ChatGPT, Codex, and DALL E. Therefore, OpenAI can also take a share of Microsoft's revenue from selling OpenAI models, but this is lower than the share that OpenAI earns from directly selling its own models.
The latest annualized revenue of $1.6 billion means that OpenAI can generate $130 million in monthly revenue through the three main methods mentioned above.
According to Bloomberg earlier this month, OpenAI is beginning discussions to raise a new round of financing, with a valuation of over $100 billion, which is 62 times its current annual revenue.
It is worth mentioning that OpenAI's most important competitor, Anthropic, supported by Amazon and Google, was also exposed today regarding its financial progress. According to sources familiar with the situation, by the end of 2024, Antiopic's annual revenue will exceed $850 million, far exceeding previous market estimates. However, compared to OpenAI, which has already reached $1.6 billion this year, it has been significantly reduced.
According to The Information, OpenAI's management believes that by the end of 2024, OpenAI's annualized revenue can reach $5 billion. From this number, it appears that the hope of Anthropic catching up with OpenAI is even more slim.
Currently, the US market no longer requires excessive investment at the level of basic large models, but top capital is still converging towards top companies. According to the statistics of COATUE, a US investment firm, as of November 2023, venture capital firms have invested nearly $30 billion in the AI field, of which about 60% have invested in large language model startups such as OpenAI and Antitopic, about 20% have invested in the infrastructure that supports and delivers these models, and only about 17% have invested in AI application companies.