OpenAI Killed Sora: Is The Bubble Bursting?
Yes, and no.
Sora, OpenAI’s text-to-video AI model, was the pinnacle of AI slop. As soon as Altman launched it in late 2024, the internet was flooded with a tsunami of objectively bad, utterly nonsensical, and often legally dubious 20-second videos that fell firmly in the uncanny valley of not-real-enough. Yet, AI glazers and tech bro fanatics still ran around like headless chickens in response, claiming this technology was enough to make Hollywood obsolete. The mismatch between reality and speculation had never been wider. Now, a little more than a year later, OpenAI has shut down the Sora model and its TikTok clone app. Not only that, but Disney has cancelled its $1 billion investment in OpenAI to licence its IP for Sora to use. So I guess Hollywood is safe for another day — what a shocker! But many have pointed to Sora’s demise as a sign that the AI bubble is bursting. So, is it? Well, yes and no. Let me explain.
We first need to ask why OpenAI killed Sora, as that is the key to this entire ordeal.
Why?
In short, money. Sora was nothing more than a liability and a cash bonfire.
Let’s start with the fact that AI video isn’t very useful for commercial (i.e., paid) purposes. Its lack of consistency between clips, total lack of emotional nuance, distracting levels of uncanny valley, awful artifacting, and noticeably inconsistent depth and placement make AI-generated video functionally useless for commercial purposes. So, Sora was never going to land any big-paying customers and would have to rely more on “hobbyist” use.
This is a problem because AI video models are insanely expensive to operate. One study found that a ten-second AI-generated video uses 30 times more energy than an AI-generated image, and that AI-generated image uses 2,000 times more energy than typical text generation. Indeed, one analyst has estimated that Sora cost OpenAI $15 million a day to operate! Yet, because Sora was only ever really used by people “playing around,” it brought in virtually no revenue.
So, this product cost OpenAI billions of dollars a year to operate, yet it produced practically zero revenue because it was only really useful for making slop videos.
There is only one possible justification for keeping such a profound waste of money on the books: social impact. If Sora had a huge social impact and functioned as incredible PR for OpenAI, they could justify keeping the service going.
But this is where the Sora 2 app comes in, because, despite the hype, it was a total failure.
