From: allin
OpenAI has demonstrated significant revenue growth, alongside discussions about its business model and market position.
Rapid Revenue Growth
OpenAI has reportedly reached a run rate of 28 million [00:50:55].
At an approximate valuation of $80 billion from secondary share sales [00:51:39], this implies the company is trading at roughly 25 times its forward revenue, a multiple similar to Nvidia [00:51:44].
Business Model Overview
OpenAI’s revenue is generated through two main components:
- Business-to-Consumer (B2C) Subscriptions: This involves a consumer subscription plan, typically costing around $20 per month [00:53:21].
- Business-to-Business (B2B) API Products: These products are sold to developers [00:54:07].
Concerns exist regarding the sustainability of the B2C model due to historically high churn rates, which can range from 5% to 10% monthly, translating to about 50% annually [00:54:29]. In contrast, the B2B segment is seen as the future value driver, with potential for 120% to 150% year-over-year expansion from existing customers [00:54:51].
[!INFO] Enterprise customers also utilize the consumer product but pay a corporate license, enabling features like consolidated searches and shared data [01:01:48].
The majority of OpenAI’s current revenue is believed to be consumer-oriented [00:56:51].
Market Position and Sustainability
OpenAI is recognized for creating “awesome products,” with ChatGPT 4o being highlighted as their leading language model [00:52:54]. However, other LLMs are “catching up” [00:54:46].
There are observations that traffic to ChatGPT in the United States decreased by 25-26% since the end of the school year, suggesting a reliance on young users for academic purposes [00:57:02]. Globally, however, traffic is still growing [00:57:23]. A key challenge is the Average Revenue Per User (ARPU) problem, where users in the U.S. are significantly more lucrative (potentially triple digits annually) compared to international markets (single or double digits) [00:57:38].
Despite rapid growth, enterprise-level adoption for production workloads using AI is still in its early stages, primarily in “experimentation” phases [00:55:35]. This is due to the probabilistic nature of AI models, which can lead to “not very good” results when stringing together multiple elements [00:56:11].
Many large enterprises are seeking open-source solutions to avoid proprietary models [00:58:40], with 80% expressing this preference [00:58:40]. Open-source models like Llama can achieve 85-90% of the desired functionality at a lower cost [00:59:36], allowing companies to tune and integrate them with their own data [00:59:47].
[!WARNING] The partnership between Apple and OpenAI regarding Apple Intelligence integration raises privacy concerns, given Apple’s historical stance on user privacy and OpenAI’s access to user data at the operating system level [00:48:00].