Why OpenAI decided to close Sora

It was supposed to be one of the flagship projects. The muscular demonstration of how perfect the algorithms built were. Instead, some two years after its launch, OpenAI has decided to close Sora. The video generation platform from text, launched in 2024, which you probably came across on social media by swiping some amazing videos, had attracted attention since its launch for its ability to produce realistic videos from textual prompts (i.e. text with a few well-written requests). And it must also be said that within a few months it had become one of the most watched products in the industry, not least because of the implications on the creative industry.

OpenAI, however, decided to resign the application. A heavy choice, from the point of view of vision and business, but certainly well-considered by Sam Altman's company.

Apparently, the motivations behind this choice are manifold. Starting with the computational cost. Generating videos with AI, in fact, requires much more resources than text and images. And ChatGPT's parent company would thus have chosen to reallocate computing capacity towards products with more immediate returns, in particular enterprise services and infrastructure.

The other big unknown about Sora was revenue. The platform had not yet found a balance between usage and monetisation. In other words: revenues did not cover operating costs, in a context where companies in the sector are increasing investments to support increasingly complex models.

But it seems that there is also a league issue behind this decision. In recent months, in fact, various newspapers such as the Guardian and the New York Times had highlighted the growing tensions on the copyright and data use front. A fragile balance that in the case of video is even more subtle: visual similarities, use of faces without authorisation, content created by imagination (therefore of any kind). This has contributed to making large-scale distribution more complex.