🤷♂️ Sam Altman Explains Why Sora Was Shut Down
Sam Altman, head of OpenAI, has for the first time подробно explained why the company made the radical decision to shut down Sora—its ambitious video generation project.
According to Altman, the main reason was a critical shortage of computing power. The company had to choose: continue developing energy-intensive video generation or allocate all resources to building next-generation AI “researcher” models. The choice was made in favor of smarter models.
Key points from Altman’s statement:
- Focus on priorities. OpenAI has previously shut down successful направления (such as its robotics division) to avoid spreading resources too thin.
- Shift in strategy. Sora was considered as part of ChatGPT, but leadership feared that chasing video views could create “misaligned incentives” for the company’s development.
- Deal with Disney. Altman personally informed Bob Iger about the project’s cancellation, admitting that disappointing such a partner is “always incredibly sad.”
Despite the collapse of the “billion-dollar deal,” OpenAI does not intend to fully sever ties with The Walt Disney Company. Altman emphasized that both companies continue “intense work” to find a collaboration format that would allow Disney to use OpenAI technologies to create “something amazing,” but without direct involvement of Sora.
At The Walt Disney Company, readiness for “constructive collaboration” was confirmed, with the focus now shifting to API usage and integrating ChatGPT into the studio’s internal processes.
It is worth noting that Sora was officially discontinued in early 2026, which came as a shock to the industry. Along with it, a $1 billion deal with The Walt Disney Company collapsed, which was supposed to give neural networks access to archives of animated films and movies.
This is not the first time OpenAI has sacrificed promising projects: the company previously shut down its robotics division and paused several ambitious reinforcement learning research initiatives, consistently citing a shortage of computing hardware needed to advance language models.