Why OpenAI really shut down Sora
Robert Way / Getty Images
One million dollars a day – that was the cost for OpenAI to maintain Sora, while the number of users plummeted from one million to just 500,000. The decision to shut down the video generation tool only six months after its high-profile debut sparked a wave of speculation regarding a potential data leak; however, the reality proved to be far more mundane. An investigation by the WSJ reveals that the project became a financial trap that posed a real threat to the company's position in the global AI arms race. The primary issue turned out to be astronomical operating costs coupled with a lack of sustained audience interest. Video generation requires immense computing power, and every user creating fantastic scenes featuring their own face exploited scarce AI chip resources that OpenAI could have utilized to develop more profitable models. For creators and creative industry professionals, this is a clear signal that the era of free or low-cost, high-quality text-to-video tools may arrive much later than anticipated. The market is currently verifying whether spectacular visual capabilities go hand in hand with profitability, as the priority for tech giants shifts toward optimizing hardware resources instead of maintaining impressive but loss-making experiments. Concentrating computing power on projects with a larger scale of utility is currently the only way to maintain the pace of innovation.
The decision by OpenAI to shut down Sora just six months after its high-profile public debut has triggered a wave of speculation in the tech industry. The tool, which was supposed to revolutionize the AI-generated video market, disappeared from the radar as quickly as it appeared. Initial conspiracy theories suggested that the project was merely a sophisticated biometric data collection operation, as the application encouraged users to upload images of their own faces to personalize generated content. However, the reality revealed by The Wall Street Journal investigation is much more prosaic, yet brutal for the image of the AI leader.
It turns out that Sora became a massive financial burden for Sam Altman's company, which at one point began to threaten the organization's strategic goals. In a world where every unit of computing power is worth its weight in gold, OpenAI had to make a cold calculation. Instead of building a new era of digital cinematography, the company faced the problem of "burning through" vast resources on a product that failed to maintain the interest of a mass audience. This is a classic case of technology that outpaced its economic viability.
A million dollars a day for an empty cinema
The numbers behind the fall of Sora are ruthless. After a spectacular start, when the number of users worldwide peaked at approximately one million people, interest began to drop sharply. In a short time, the base of active creators shrank to below 500,000. In the Software as a Service (SaaS) industry, such a churn of users during the growth phase is an alarm signal that cannot be ignored, especially with such high operating costs.
Read also
Keeping Sora alive cost OpenAI about one million dollars a day. Importantly, these huge sums did not result from intensive use by a loyal community, but from the very nature of AI video generation. This process is incomparably more computationally demanding than generating text or static images. Every user who wanted to see themselves in a fantastic setting consumed scarce resources of NVIDIA chips and other AI accelerators that could have been training next-generation models like GPT-5 at the same time.
Priorities in the AI arms race
In the face of growing competition from Anthropic, Google, and Meta, OpenAI could not afford to waste computing power. The AI market is currently at a point where access to graphics processing units (GPUs) defines winners and losers. Every server dedicated to supporting Sora was a server taken away from teams working on the company's technological foundations. Internal analyses showed that maintaining an unprofitable and declining video tool slowed down the development of key language models.
The "all or nothing" strategy in training increasingly powerful Artificial General Intelligence (AGI) systems forced the OpenAI board to make radical cuts. Sora, despite its visual glitz, proved to be a niche product that could not find a sustainable commercial application beyond short-lived fascination with the novelty. In the world of creative technologies, the barrier to entry turned out to be not only the cost but also the time required to generate materials, which was often unacceptable for professional creators.
A lesson in generative economics
The downfall of Sora sheds new light on the problem of scalability for multimedia models. Although this technology can generate photorealistic footage, its "appetite" for electricity and silicon is currently too great for a business model based on broad public access. OpenAI likely miscalculated estimates regarding algorithm optimization — it was assumed that over time the cost of generating a video frame would drop, similar to what happened with DALL-E. However, this did not happen to a degree sufficient to save the project.
- Operating costs: $1 million per day with a declining number of users.
- Hardware resources: Necessity to allocate AI chips to more important research projects.
- User retention: Drop from 1 million to less than 500,000 in a few months.
- Competition: Pressure for rapid development of text and multimodal models.
Instead of a revolution, we received a costly experiment that showed the limits of current cloud infrastructure. Sora was not a "data trap," but a victim of its own complexity and the ruthless economics of scale that promotes efficient solutions rather than just effective ones.
The closure of Sora is a signal to the entire sector: the era of free or cheap experiments with generative video on a mass scale is coming to an end, giving way to a hard fight for access to computing power.
OpenAI's withdrawal from the video market does not mean the company is completely abandoning the field. The technology will likely return in the future, but in a completely different form — perhaps as a closed API for major film studios or an integrated element of a more comprehensive model that will be much better optimized for resource consumption. For now, however, the priority remains dominance in the foundation model segment, where the company's survival in the global race for artificial intelligence supremacy is at stake. Sora was a luxury that OpenAI, despite billions in investment, simply could no longer afford.
