$15 Million a Day to Run. Zero Path to Profit. The Sora Shutdown Was Inevitable.

For six months. That was the duration of Sora’s existence as a stand-alone application before OpenAI silently started turning off the lights and posted a farewell message on X. “To everyone who created with Sora, shared it, and built community around it: thank you,” the business wrote in a note. It was the type of statement that reads politely at first glance and a little hollow at second. One of the most significant AI products in recent memory was supposed to be Sora. Rather, it turned into a case study about the gap between what technology can accomplish and what a business can truly afford to continue operating.
No executive statement can convey the story as quickly as the numbers. OpenAI’s compute costs for each ten-second video produced by Sora came to about $130. Not $130 a month for each user. Not $130 a day. according to the video. After the app was released in late 2024, millions of users created content every day, with over a million downloads in the first five days alone. As a result, those expenses quickly increased. An estimated $15 million was spent on daily operations. That amount is close to $5.4 billion annually. To put it mildly, the economics were not working for a product that brought in about $1.4 million in net in-app revenue.

It’s important to keep in mind how the Sora launch appeared at the time. The clips, which were incredibly realistic videos put together from text prompts and shared on social media with such speed that the rounds seemed inevitable, were truly remarkable. Animals acting ridiculously. cinematic sequences that appeared to have been filmed on location. The online reaction had that unique quality of widespread awe that only occurs once every technological cycle, right before novelty turns into expectation. Sora rose to the top of the Apple App Store in a matter of days. It was obvious that people wanted to use what OpenAI had created. The issue was that “willing to pay for it” and “wanting to use it” proved to be quite different demographics.

As this story develops, it seems as though the warning signs were present before the announcement. According to reports, Fidji Simo, CEO of applications at OpenAI, told staff members that the company “cannot miss this moment because we are distracted by side quests” and that they should concentrate on productivity tools, particularly on the business side. This is the language used by a company to justify its product line in advance of something important, which in this case is an expected initial public offering (IPO) that may take place before the end of 2026. A cogent story about the source and destination of the funds is necessary for going public. Contrary to that story is a product that burns $15 million every day and has no obvious way to make money.
Disney comes next. Alongside Sora’s closure, the $1 billion licensing agreement that had been announced in December, giving OpenAI access to over 200 iconic characters for use in video creation, fell through. It’s amazing to hear that Disney was taken aback by the announcement given the size of the deal. It’s unclear if the surprise is the result of a communication breakdown, an abrupt acceleration of OpenAI’s internal decision-making, or just the chaos that usually follows significant strategic shifts. It is evident that SAG-AFTRA, which had spent years forcing consent and compensation clauses into its agreements with Hollywood studios, can use this result as proof that its stances weren’t irrational. The union’s control over the use of licensed characters led to moral and legal dilemmas that exacerbated the expenses Sora was already finding difficult to control.
In addition to describing Sora to the BBC as “a resource black hole” with “limited monetization,” Forrester analyst Thomas Husson also pointed out that the platform had trouble stopping the production of realistic misinformation and non-consensual imagery. These are serious issues in a pre-IPO setting where regulatory and reputational risk are partially reflected in a company’s valuation. AI video production operates in a particularly delicate legal environment, where issues pertaining to copyright, likeness, and consent are still being addressed by courts rather than resolved by precedent. Apparently, OpenAI made the decision not to wait for those responses.
Sora’s closure might be viewed as a single example of a larger correction occurring throughout the AI sector. The “move fast and demo everything” era, which gave rise to some truly remarkable technology and some truly unsustainable business models, appears to be giving way to something more limited and concentrated on demonstrating the existence of revenue. According to reports, ChatGPT made about $1.9 billion in in-app revenue while Sora made only $1.4 million. OpenAI’s priorities are better explained by that comparison than by any strategic memo.
The options are already in circulation. The AI video generation market isn’t shrinking because Sora left it, as evidenced by Google’s Veo, Runway, Pika, and Luma AI’s Dream Machine. If anything, other businesses are still navigating and attempting to resolve the compute and rights issues that caused Sora to fail. It’s genuinely unclear if any of them will discover a model that works. Sora’s departure demonstrated that creating videos based on text prompts is an extremely costly process and that enthusiasm, even viral, record-breaking enthusiasm, is not the same as a business.