OpenAI projects to reach $200 billion in revenue by 2030, a figure comparable to Microsoft's current $245 billion. This is a colossal leap from the $12 billion expected this year, as the company plans to multiply its revenue by 17 in less than five years. For perspective, Apple took forty years to reach that level; Google, twenty. OpenAI aims to achieve this in just fifteen years, with a key detail: until 2022, its revenue did not exceed $100 million. The growth chart published by The Information reveals more than ambition. Revenue skyrockets exponentially, while computing costs—both for training and running models—grow much more linearly. This equation only holds if OpenAI transcends its current subscription model for LLMs at $20 a month. The real question is not whether they can multiply their revenue, but what they need to create to justify such a valuation. The key lies in agents, but not in the way we imagine. OpenAI is not looking to sell you a more advanced ChatGPT, but to replace entire departments. The model is already emerging with Deep Research: not charging per query, but per work done. If a report that previously required three analysts for a week can now be generated by an agent in minutes, supervised by just one person, its value is not that of a subscription, but of the salaries it replaces: tens of thousands of dollars. Multiplied by every department in every Fortune 500 company, the $200 billion no longer seems like science fiction. It may even be an underestimate. But here lies the paradox: to capture that value, OpenAI needs its models to be irreplaceable. However, competition—Claude, Gemini, DeepSeek—narrows the gap every month. AI is becoming a commodity. How can they justify premium prices when their product increasingly resembles a basic utility? OpenAI's answer is speed. It’s the strategy of Uber or Amazon: invest huge sums to gain market share, trusting that when profitability arrives, only they will remain standing. The alternative plan lies in vertical applications. They will not sell generic AI, but specific solutions: complete customer service systems, educational platforms, legal copilots. Each vertical is a multi-billion dollar market. Microsoft 365, for example, generates nearly $100 billion annually. The enterprise software market moves close to a trillion a year. If OpenAI manages to replace just 20% of that market with intelligent agents, it will reach its goal without needing to invent anything radically new; it just needs to make the existing obsolete. OpenAI's true bet is not just technological, but temporal. It is buying time with massive investments in computing, betting that AGI—or something sufficiently similar—arrives before resources run out. If they succeed, the $200 billion will be just the beginning. If they fail, we will have witnessed the largest tech bubble in history. What’s fascinating is that not only OpenAI is betting on this future: Microsoft, Oracle, SoftBank, and even the U.S. government seem convinced they can achieve it.