The article argues that OpenAI is presenting itself as many companies at once without a coherent plan, citing a flurry of reported initiatives that range from a new social feed of generative video called Sora 2 to a potential productivity suite aimed at Microsoft’s turf. Other purported efforts include an Artificial Intelligence powered hiring platform targeted for mid-2026, advertising inside ChatGPT by 2026, a possible move into selling infrastructure services later, in-house Artificial Intelligence chips with Broadcom slated for 2026 but for internal use, consumer hardware by late 2026 or early 2027, and even a browser. The author frames many of these as strategic leaks designed to bolster valuation and facilitate massive future fundraising on a trillion-dollar scale.
At the core, the piece contends that OpenAI lacks focus and that its flagship model update, GPT-5, was underwhelming and more expensive to operate than its predecessor due to how it processes prompts. Citing projections reported by The Information, the author says ChatGPT is expected to remain the dominant revenue driver until at least 2027, when new “agents” and monetization for free users are supposed to contribute meaningfully. The article questions whether OpenAI is a hardware company, software vendor, ads platform or cloud provider, noting that even ideas like certifying Artificial Intelligence experts are floated while the company’s identity remains unclear.
Financially, the essay characterizes OpenAI as a standard software business that makes most of its money from ChatGPT subscriptions. It references 20 million paid subscribers as of April and 5 million business subscribers as of August, including 500,000 seats from the Cal State University system. The author says the company loses large amounts of money and that API revenue appears to be a very small share in 2025, with the company’s “Operator” agent described as barely functional. That dynamic, the piece argues, makes OpenAI resemble any other Artificial Intelligence startup trying to bolt large language models onto products while struggling to monetize.
Beyond business execution, the article points to foundational limits in large language models, noting that “hallucinations” are described as mathematically inevitable by OpenAI’s own research. It further claims that OpenAI’s growth is slowing, its models are increasingly commoditized, and the broader generative Artificial Intelligence narrative has cooled. According to The Information, OpenAI spent roughly 150 percent of its first-half 2025 revenue on research and development, producing the muted GPT-5 release and Sora 2. The author estimates that Sora 2 carries high per-video generation costs based on published cloud rates for the earlier Sora model and questions whether those economics are sustainable.
