Recently, OpenAI's CFO, Sarah Friar, penned an article detailing ChatGPT's remarkable evolution over the past three years. From a novelty that piqued curiosity, ChatGPT has now morphed into an indispensable infrastructure, shaping the daily work and lives of hundreds of millions worldwide. For the first time, Friar comprehensively unveiled OpenAI's business rationale and growth metrics, which are intricately linked to compute power and priced in accordance with the 'intellectual value'.
The article highlights that ChatGPT was initially conceived to gauge the impact of placing cutting-edge intelligence in the hands of the general public. However, its subsequent widespread adoption surpassed all expectations. Students, parents, writers, and others harnessed its capabilities to tackle diverse challenges, transforming it into a personal 'asset'. This asset swiftly transitioned into the professional realm, with ChatGPT seamlessly integrating into daily work routines and evolving into an infrastructure that empowers users to 'create more, decide faster, and operate at a superior level'.
OpenAI adheres to the principle of synchronizing its business model expansion with the escalating value of intelligence. It has launched subscription services, workplace subscription products, and introduced usage-based billing to construct a platform-centric business. Recently, OpenAI has extended its business logic to the advertising domain, offering clearly labeled and genuinely beneficial options when users are on the verge of making decisions.
In terms of usage statistics, ChatGPT's weekly and daily active users continue to set new records. Compute investments propel advancements in frontier research and model capabilities, with more robust models translating into superior product experiences and wider platform adoption. This adoption, in turn, generates revenue, which is reinvested into compute and innovation, creating a virtuous cycle. From 2023 to 2025, compute power witnessed an average annual surge of approximately threefold, culminating in an overall growth of about 9.5 times. The revenue trajectory mirrored this growth, with an average annual increase of roughly three times during the same period, soaring from $2 billion in annual recurring revenue in 2023 to $6 billion in 2024, and surpassing $20 billion in 2025. Sarah Friar characterized this as 'unprecedented growth on such a massive scale' and posited that had more compute power been accessible, the pace of customer adoption and commercial realization would have been even swifter. In her estimation, 'compute power represents the most scarce resource in the contemporary artificial intelligence landscape'.
Three years ago, OpenAI relied almost exclusively on a single compute provider. Today, it operates within a diversified ecosystem, collaborating with multiple suppliers to ensure 'compute certainty'. When training frontier models, OpenAI employs top-tier hardware during critical capability stages and prioritizes cost-effectiveness when handling large-scale inference workloads, making artificial intelligence applicable to daily workflows.
Leveraging compute power, OpenAI has forged an integrated product platform encompassing text, images, voice, code, and APIs. The next frontier is 'intelligent agents and workflow automation'. As these systems transition from 'novelty' to 'habit', their usage will deepen and become more sustained, underpinning long-term investment. Looking ahead, as intelligence permeates more domains, novel economic models will continue to emerge. Sarah Friar notes that this trajectory mirrors the evolutionary path of the internet industry, and 'intelligence will follow a similar course'.
OpenAI navigates business growth contradictions through strategies such as maintaining a 'light asset' balance sheet, substituting proprietary full-stack solutions with partnerships, and preserving contractual flexibility. In 2026, OpenAI's core focus will shift to 'practical adoption', with the paramount task of bridging the gap between 'what artificial intelligence can already accomplish today' and 'what individuals, businesses, and countries actually utilize in their daily lives'.
