From Limelight to Abrupt Exit: Why Sora Became OpenAI's "Abandoned Asset"

OpenAI once positioned its video generation tool Sora as its most significant consumer product after ChatGPT, but has now abruptly halted it less than six months after its public release. This outcome suggests that in the AI arena, where compute power is scarce and competition is intensifying, products that burn cash without a clear path to monetization have no room to survive.

The immediate trigger for Sora's shutdown was its grim commercial ledger— global active users plummeted from a peak of approximately one million to fewer than 500,000, with daily operating losses reaching about $1 million. Meanwhile, OpenAI is accelerating the allocation of compute power for a new model codenamed "Spud" to support the development of programming and enterprise service products, while Sora was consuming an excessive amount of this most scarce resource.

This decision dealt a direct blow to OpenAI's core partners. Disney had previously signed a multi-year licensing agreement with OpenAI, pledging to invest $1 billion and opening up over 200 of its IP characters for Sora's use. Many Disney executives only learned of the shutdown decision less than an hour before the public announcement; the $1 billion investment has yet to materialize, and the partnership has effectively stalled.

OpenAI stated it will redirect the freed-up compute power toward productivity tools to catch up with its competitor Anthropic, which has gained a first-mover advantage in the enterprise market. This pivot indicates that as the critical IPO juncture approaches, commercial monetization capability has become the primary standard for the company's resource allocation.

A Fleeting Phenomenon: Growth Pains Behind the Hype

Sora's creation stemmed from the academic ideals of two doctoral students from the University of California, Berkeley. Tim Brooks and Bill Peebles joined OpenAI in early 2023, dedicated to building an AI model capable of generating high-quality videos from text and simulating the physical world. In February 2024, they named the system "Sora," Japanese for "sky," and unveiled it to the public for the first time. Sora stunned the industry with lifelike videos such as a mammoth traversing a snowy landscape and a fashionable woman walking through the neon streets of Tokyo. Sam Altman immediately invited users on X to submit text prompts to showcase its generation capabilities.

In December of the same year, OpenAI officially launched Sora as a consumer-grade application for the public. Within a week of its launch, the application quickly climbed to the top of the App Store. Users could generate a ten-second short video in minutes by simply inputting prompts; additionally, it supported uploading personal faces, allowing users to star in various fantastical scenarios. Altman himself contributed his image, sparking a wave of user experimentation.

However, the popularity waned just as rapidly. After reaching a global peak of about one million users, the number continued to decline, shrinking to fewer than 500,000 in the following months. According to data from analysis firm Similarweb, usage had largely stagnated by the end of the year. Altman had once likened Sora's launch to the historic moment of ChatGPT's debut, but the application ultimately failed to realize its creators' vision—The Wall Street Journal described its actual performance as "more AI dross than AI magic."

Compute Black Hole: The $1 Million-a-Day Price Tag

The high operating cost of Sora stems from the technical characteristics of video generation models themselves. Unlike language models that learn from text, video models need to understand and reconstruct complete dynamic scenes, making training and inference costs far higher than text-based products. Every user embedding their face into a World War II newsreel or a Hollywood car chase scene consumes a finite amount of AI compute power. According to media reports citing individuals familiar with the matter, Sora's daily operating losses reached approximately $1 million.

OpenAI has an internal dashboard tracking compute allocation for various teams. Some employees expressed surprise at the proportion of compute power allocated to the Sora team—a video generation tool that neither generated substantial revenue nor improved language model capabilities raised internal questions about resource allocation.

External competitive pressure made this issue even more urgent. Google's Gemini gained widespread users on the consumer end, and Anthropic's code tool Claude Code rapidly captured the Silicon Valley software engineer community with its highly autonomous programming capabilities, catching OpenAI off guard. OpenAI hastily released a new version of its own programming product, Codex, but still struggled to close the gap.

The new model, codenamed "Spud," urgently required more compute power, and the company also planned to train a separate new model for video generation within ChatGPT. After calculating the costs, OpenAI ultimately decided to cancel this training plan and completely shut down Sora.

Meta's Poaching and Internal Silos

Sora nearly met its end prematurely due to a talent war. In the spring of 2025, Meta CEO Mark Zuckerberg launched a large-scale recruitment drive targeting OpenAI, personally contacting dozens of top researchers with high salaries to join his AI lab. Sora co-founder Bill Peebles was on the list and seriously considered the offer. According to media reports citing insiders, OpenAI eventually retained Peebles with a salary increase and later expanded his responsibilities on the Sora project, tasking him with leading the training of the next-generation video generation model and the development of the consumer application.

However, within the company, Sora maintained a highly closed operational state. The project belonged to the World Simulation team led by Aditya Ramesh and operated independently from the core research team responsible for ChatGPT, with its progress kept highly confidential from other departments. Some former employees described Sora as a "company within a company." This isolated operating model, to some extent, prevented Sora's strategic value from gaining broader internal recognition and placed it in a relatively vulnerable position in the company's resource competition.

Disney's Shattered Dream: A $1 Billion Investment Evaporates

Sora's most representative partnership came to an equally sudden end.

In December 2024, OpenAI and Disney announced a multi-year licensing agreement covering over 200 renowned IP characters from Marvel, Pixar, and others. Disney simultaneously committed to investing $1 billion in OpenAI and becoming one of its major enterprise clients. Disney CEO Bob Iger stated in an interview with CNBC that this partnership would allow Disney to participate in the rapid growth of AI and the shaping of new media entertainment forms. Notably, the day before announcing the agreement with OpenAI, Disney had sent a cease and desist letter to Google for copyright infringement, a move seen as a significant signal of the company's endorsement of AI licensing business models.

In February of this year, Iger revealed on an earnings call that short videos produced using Sora would soon appear in Disney+'s vertical video feed, and negotiations were underway to introduce ChatGPT across the entire Disney organization.

However, before OpenAI announced the shutdown of Sora, many Disney executives had less than an hour's notice. The $1 billion investment has not yet been finalized, and the partnership has effectively stalled. Disney subsequently issued a statement saying: "As the AI field rapidly evolves, we respect OpenAI's decision to exit the video generation business and shift its focus to other areas. We appreciate the productive collaboration between our teams." Under the leadership of new CEO Josh D'Amaro, Disney is currently actively consulting with over a dozen partners on plans to introduce other AI tools.

Betting on "Super Apps": Pragmatic Logic Takes Over

Sora's closure is not merely a product failure; it is a microcosm of OpenAI's accelerated strategic focus contraction on the eve of its IPO.

Altman characterized the decision as a "difficult but necessary sacrifice for the company's overall goals" in an internal letter to employees, expressing gratitude for their willingness to make "tough trade-offs." An OpenAI spokesperson stated that the company is engaged in "ruthless prioritization" of compute resources based on the principle of maximizing long-term economic value, adding that "this focus allows us to grow and innovate faster, and serve enterprises and developers more efficiently."

The company's current focus is shifting toward a planned "super application" that will integrate so-called "agent" AI tools, enabling them to autonomously perform tasks for users such as writing code, analyzing data, and booking travel. Such productivity-oriented products are rapidly gaining traction among enterprises and developers, but OpenAI has lagged behind Anthropic in this market. Altman stated that the original Sora team's focus will shift to areas with greater long-term potential, such as robotics.

For OpenAI, Sora's journey represents a costly strategic misjudgment. This product, once envisioned by Altman to reshape popular culture and open new revenue streams, ultimately faltered due to a lack of a viable business model and flawed resource allocation. At this critical juncture focused on profitability, it became a burden that had to be shed.