Why OpenAI killed Sora

Compute costs stronger than artistic vision

The main reason for killing off **Sora** turned out to be brutal economic realities. Since its debut, this model consumed gigantic resources of computing power (**compute**) without generating a financial return that could justify such high maintenance costs. In the world of AI, computing power is a currency rarer and more valuable than the dollar, and **OpenAI** had to choose between a flashy video generator and the development of productivity tools that actually pay the bills. **Fidji Simo**, serving in the new role of CEO for AGI deployments, made the matter clear in talks with employees: the company cannot afford to be distracted by "side quests." Business efficiency and enterprise tools have become the priority. As a result, not only **Sora** suffered, but also other experimental features, such as the previously considered "adult mode" in **ChatGPT**. This decision reflects a broader trend in the industry — investors, who until now were eager to fund visionary demos, are now demanding hard evidence that the technology can earn its keep.

Lack of advantage in a crowded market

Despite a spectacular start, **Sora** quickly began to lose ground to the competition. Industry experts, including **Trevor Harries-Jones** from the **Render Network Foundation**, point to the lack of a so-called market moat. In the generative video sector, innovation is progressing so fast that users easily switch to tools offering better quality or lower costs. Models from **Google** or the Chinese **Kling** proved to be not only cheaper to operate but in many cases more technically advanced. Market statistics provided by **Sensor Tower** confirm this regression. Although **Sora** recorded a great debut in October with **4.8 million** downloads, subsequent months brought sharp declines:

• November: 6.1 million downloads (peak popularity)
• December: 3.2 million
• January: 2.1 million
• February: 1.4 million
• March: just 1.1 million (forecast for the end of the month)

What is particularly concerning for analysts is that these declines occurred at a time when the application was entering new geographical markets, which theoretically should have stimulated growth. The lack of interest from users ultimately sealed the project's fate.

The end of a billion-dollar romance with Disney

One of the most surprising effects of the new strategy is the termination of cooperation with **Disney**. The deal, worth **$1 billion**, was intended to be a milestone in the history of AI in entertainment. It assumed that characters from the **Star Wars**, **Marvel**, and **Pixar** universes would be available in the **Sora** model, and the generated content would go directly to the **Disney+** platform. The breakup occurred just three months after signing the three-year contract. Although for **OpenAI** this means losing a powerful client and brand partner, **Disney** does not intend to give up on AI. Representatives of the entertainment giant made it clear that they are open to licensing their brands to other players, such as **Runway**, **Luma**, or **Google**. This shows that the problem was not the video technology itself, but **OpenAI**'s inability to deliver a stable, profitable product within the imposed corporate rigors.

New direction: Robotics and world simulation

Closing the consumer application does not mean that the research team's work went to waste. **OpenAI** spokesperson **Kayla Wood** declared that resources will be redirected to research on **world simulation**. This is of key importance for the development of robotics, where AI must learn physics and interaction with the real environment to help humans with physical tasks.

"We have decided to withdraw Sora from the consumer app and API. The research team is now focusing on world simulation research to support robotics in solving real-world tasks," Kayla Wood reported.

This shift in focus makes sense in the context of the rivalry with **Anthropic**. While **Sora** generated "AI slop" (low-quality content flooding the internet), **Anthropic** was building a reputation as a leader in coding tools and enterprise solutions. **OpenAI** must now prove it can deliver high-value-add tools, especially in light of the planned **IPO** this year. Giving up on the **Sora** model is a painful but necessary lesson in humility for **Sam Altman** and his team. It shows that in the AI arms race, the winner is not the one who creates the most impressive demo for Twitter, but the one who harnesses infrastructure costs and finds a real application for the technology in business. However, the legacy of the **Sora** model will remain with us in the form of lasting uncertainty regarding the authenticity of video materials online — a problem that no application closure will solve. The company's future now depends on whether it can turn billions of dollars from investors into real profit before the bubble of expectations surrounding generative artificial intelligence bursts.