OpenAI once positioned its video generation tool Sora as its most important consumer product following ChatGPT. However, it was abruptly discontinued less than six months after its public release. This outcome suggests that in the fiercely competitive AI landscape, where computing power is scarce, products that burn cash without generating revenue have little room to survive.
The immediate trigger for Sora's shutdown was a grim financial reality. Its global active user base plummeted from a peak of approximately one million to under 500,000, with daily operational losses reaching around one million dollars. Concurrently, OpenAI is prioritizing computing resources for a new model, codenamed "Spud," to bolster its programming and enterprise service products. Sora was consuming an excessive amount of this most scarce resource.
This decision directly impacted OpenAI's core partner,
OpenAI stated it will reallocate the freed computing power to productivity tools, aiming to catch up with competitors like Anthropic, which has gained an early lead in the enterprise market. This strategic pivot indicates that, as the company approaches a potential IPO, commercial viability has become the primary criterion for resource allocation.
**A Fleeting Success: Growth Challenges Behind the Hype**
Sora originated from the academic vision of two UC Berkeley PhDs, Tim Brooks and Bill Peebles, who joined OpenAI in early 2023. They aimed to build an AI model capable of generating high-quality video from text, thereby simulating the physical world. In February 2024, they named the system Sora, after the Japanese word for "sky," and unveiled it to the public. It stunned the industry with realistic videos, such as a woolly mammoth traversing a snowy plain and a stylish woman walking through Tokyo's neon-lit streets. Sam Altman subsequently invited users on the X platform to submit text prompts to showcase its capabilities.
In December of the same year, OpenAI officially launched the Sora consumer application. Within a week, it topped the App Store charts. Users could input a prompt and receive a ten-second video clip within minutes. The ability to upload one's own face allowed users to insert themselves into various fantastical scenarios, a feature Altman himself demonstrated, sparking widespread user experimentation.
However, the hype faded quickly. After reaching a peak of roughly one million users, the global user count declined steadily, shrinking to under 500,000 within months. Data from analytics firm Similarweb indicated usage had plateaued by year-end. Altman had compared Sora's launch to the historic release of ChatGPT, but the application ultimately fell short of its creators' vision, with one publication describing its output as "more AI dross than AI magic."
**A Compute Black Hole: The Million-Dollar Daily Cost**
Sora's high operational costs stem from the inherent technical demands of video generation models. Unlike language models that learn from text, video models must understand and reconstruct complete dynamic scenes, making training and inference costs significantly higher than text-based products. Every user embedding their face into a WWII newsreel or a Hollywood car chase scene consumed a portion of finite AI computing power. According to reports citing informed sources, Sora's daily operational losses were approximately one million dollars.
Internally, OpenAI maintains a dashboard tracking compute allocation across teams. Some employees were surprised by the proportion of resources granted to the Sora team—a video tool that generated negligible revenue and did not enhance core language model capabilities, prompting internal questions about this resource distribution.
External competitive pressures made this issue more urgent. Google's Gemini gained a broad consumer user base, while Anthropic's Claude Code, with its highly autonomous programming capabilities, quickly captivated Silicon Valley software engineers, catching OpenAI off guard. OpenAI rushed out a new version of its own programming product, Codex, but struggled to close the gap.
The new "Spud" model urgently required more compute, and the company also planned to train a separate new model for video generation within ChatGPT. After calculating the costs, OpenAI ultimately canceled that training plan and shut down Sora entirely.
**Meta's Poaching and Internal Silos**
Sora nearly met an earlier end due to a talent war. In the spring of 2025, Meta CEO Mark Zuckerberg initiated a large-scale recruitment campaign targeting OpenAI, personally contacting dozens of top researchers with lucrative offers to join his AI lab. Sora co-founder Bill Peebles was among those approached and seriously considered the offer. Reports indicate OpenAI retained Peebles with a salary increase and later expanded his responsibilities, putting him in charge of training the next-generation video model and developing the consumer application.
Yet, within the company, Sora operated as a highly isolated entity. The project belonged to the World Simulation team led by Aditya Ramesh, which was separate from the core research team responsible for ChatGPT's language models. Its progress was kept highly confidential from other departments. Some former employees described Sora as "a company within a company." This siloed operation likely hindered broader internal recognition of Sora's strategic value, leaving it vulnerable in internal resource competitions.
**Disney's Dream Shattered: A One Billion Dollar Investment Evaporates**
Sora's most prominent partnership ended just as abruptly.
In December 2024, OpenAI and
In February, Iger revealed on an earnings call that short videos created with Sora were slated to appear in Disney+'s vertical video feed, and negotiations were underway to introduce ChatGPT across
However, many
**Betting on a 'Super App': Pragmatism Takes Over**
Sora's closure is not merely a product failure but a reflection of OpenAI's rapidly narrowing strategic focus ahead of a potential IPO.
In an internal memo to staff, Altman characterized the decision as a "difficult but necessary sacrifice for the company's broader goals," expressing appreciation for employees willing to make "hard trade-offs." An OpenAI spokesperson stated the company is conducting "ruthless prioritization" of compute resources based on maximizing long-term economic value, claiming "this focus allows us to grow and innovate faster, and serve enterprises and developers more effectively."
The company's current focus is shifting to a planned "super app" that will integrate so-called "agent" AI tools capable of autonomously performing tasks like writing code, analyzing data, and booking travel for users. Such productivity-oriented products are gaining rapid traction with enterprises and developers, a market where OpenAI currently trails Anthropic. Altman indicated the former Sora team will now shift its focus to areas with longer-term potential, such as robotics.
For OpenAI, Sora's journey represents a costly strategic miscalculation. A product that Altman once hoped would reshape popular culture and open new revenue streams ultimately failed due to an absent business model and misallocated resources, becoming a burden that had to be shed as the company sharpens its focus on profitability.
Comments