The world of generative artificial intelligence dominates both media and financial headlines. In this rapidly expanding context, OpenAI embodies both spectacular innovation and a striking economic paradox.
ChatGPT, its flagship product, attracts approximately 800 million unique users every week. Yet, the more its audience grows, the heavier the losses become. Behind the promises of a bright future lies a crucial question: is the current economic model sustainable?
An article from 6actualités mentioned our Yiaho platform as an example of a viable economic model, based on advertising. Could OpenAI follow the same path? Can its current economic model last, or is the AI bubble destined to burst sooner?
OpenAI: A Widening Financial Abyss
In the first half of 2025, OpenAI generated $4.3 billion in revenue, a 16% increase compared to the entirety of 2024.
A solid figure, but negligible compared to expenses: the company consumed $2.5 billion in cash (cash burn) over six months, including $6.7 billion solely in research and development (including stock-based compensation). On an accounting basis, net losses reached $13.5 billion, inflated by non-cash adjustments on convertible rights.
In the third quarter, losses exceeded $12 billion: a record that already surpasses total losses for 2024!
Expensive Queries
Each query sent to ChatGPT incurs a cost far greater than the value it generates. It’s often said that a ChatGPT query consumes a lot of water, but it also involves considerable expenses.
The servers required to run demanding models like Sora 2, as well as the salaries of top engineers, often poached from Meta or Apple, make this equation difficult to balance.
The more success increases, the more the deficit grows: this is the opposite of a traditional leverage effect!
Read more on this topic: OpenAI Snags Jony Ive, Former Apple Designer, to Revolutionize AI
Titanic Ambitions, Dizzying Figures
Sam Altman, OpenAI’s CEO, makes no secret of his aims. He plans to invest $1.4 trillion to build 30 gigawatts of data centers, equivalent to the electricity consumption of some 22.5 million American households.
In parallel, he anticipates $100 billion in annual revenue by 2027. Today, the company aims for $13 billion in revenue for 2025, while expecting a cash burn of $8.5 billion.
The gap between forecasts and current reality remains abysmal, with cumulative losses projected at $115 billion by 2029.
As of June 30, 2025, OpenAI nevertheless held $17.5 billion in cash and marketable securities. A comfortable cushion, largely fueled by Microsoft, whose $71 billion in free cash flow for fiscal year 2025 and nearly $100 billion in net income offer strategic support.
The massive integration of Copilot into the Microsoft ecosystem (Office, Azure, GitHub) could eventually transform these investments into recurring revenue. But is that enough to withstand the competition?
Monetization: The Quest for New Levers
To reverse the trend, OpenAI is exploring several avenues. Targeted advertising, for example, directly within the ChatGPT interface. Advertising on ChatGPT, an idea that often comes up in the tech sphere.
OpenAI could also explore another revenue source, with in-app purchases via partnerships with platforms like Walmart or Shopify.
But their ability to generate sufficient margins in the face of constantly rising inference costs remains uncertain. And only 5% of ChatGPT users subscribe to a paid version, despite the service’s popularity.
R&D: A Bet on the Future?
Fundamental research represents the heaviest expenditure.
Developing ever more powerful models requires colossal resources in computing, data, and talent. OpenAI is betting that these investments will lay the groundwork for long-term technological dominance, transforming current losses into exponential profits tomorrow.
Historically, tech giants like Amazon or Tesla have followed similar trajectories: years of massive losses before explosive profitability. For example, Amazon was not profitable for 9 years: from 1995 to 2002, it accumulated annual net losses. The first net profit came in 2003 with $35 million. In terms of cash flow, losses lasted until 2001, for 7 years! But in the meantime, the dot-com bubble had burst…
OpenAI is not profitable today, and there’s no indication of breaking even for several years. ChatGPT’s success paradoxically accelerates losses, while infrastructural ambitions push back the horizon of profitability.
Supported by Microsoft and driven by a long-term vision, the company has the means to endure. It remains to be seen whether the promises of 2027 are not a mirage in the desert of data centers!
Source:
