"We are saying goodbye to Sora. We know this news is disappointing. What you create with Sora is very important." From its stunning debut in February 2024, to the launch of an independent application in September 2025, and now quietly offline - Sora's life cycle is only two years, and its official launch is only half a year. Also coming to an end was Disney’s $1 billion investment cooperation agreement.
Sudden announcement of product closure
On Tuesday, local time in the United States, OpenAI issued an announcement with only a few dozen words on social media, announcing the shutdown of its Sora business. This company, a leader in the global AI industry, has decisively shut down the Vincent Video product that shocked the entire Hollywood two years ago.
How sudden was this decision? Just one day before announcing the closure of Sora, OpenAI had just published a blog post about how to safely use the tool to create content. Partner Disney said that as long as the artificial intelligence platform respects intellectual property rights, it "will continue to cooperate with the artificial intelligence platform."
OpenAI co-founder and CEO Sam Altman later wrote in an internal email that the closure of Sora was to shift strategic focus and resource optimization. He explained that video generation is extremely computationally expensive. In order to cope with competition from rivals such as Anthropic in the field of code generation and enterprise-level tools, the company decided to shift precious computing resources and talents to more productive AI agents, programming tools such as Codex, and the upcoming next-generation large model.
Altman said that OpenAI is streamlining its product line and integrating its products into a unified "super App" that combines ChatGPT, programming tools and browser functions into one. Sora, which exists as an independent social/media attribute, no longer conforms to this "productivity first" logic.
He also assured employees that Sora’s core research team will not be disbanded, but will turn to “world simulation” research, aiming to advance robotics technology to help solve practical tasks in the physical world, rather than just generating short video entertainment. However, what Altman did not mention is that the reason behind the closure of the star product Sora is that OpenAI is reducing capital consumption and actively preparing for listing.
Stunning debut that shakes up the industry
Sora's first appearance was stunning. In February 2024, OpenAI released the first Sora demonstration video, showing a fashionably dressed woman walking on the streets of Tokyo at night. The woman is wearing sunglasses, smiling confidently, and the wet street reflects neon light and shadow. The lens language is fluent, the light and shadow details are lifelike, and the camera movement is even more professional. In one fell swoop, AI-generated videos were transformed from toys to film and television levels.
At that moment, the entire Internet and film and television circles were stunned. Sora used a one-minute video to tell the world: AI has begun to understand the physical world. While Silicon Valley's venture capital community is excited, Hollywood is panicking. However, Sora was initially only trialled by a very small number of developers and leading media figures, and the general public could only watch with envy as they shared the AI video content they generated.
After a long wait of one and a half years, Sora 2.0 will be officially launched as an independent application in September 2025. Support text generation video, audio synchronization, and more accurate physical simulation. On the first day of launch, Sora shot to No. 1 on the Apple App Store's photography and video download list.
I remember that at that time, invitation codes were hard to find in the circle of technology enthusiasts, and various AI-generated short videos continued to become hot spots on the Internet. I also got the invitation code from an AI entrepreneur in Silicon Valley. I couldn’t wait to generate a video of myself completing a triathlon in Hawaii. I finally realized my dream with the help of AI.
During that time, a large number of fake, bizarre, and shocking videos appeared on social networks, most of which came from the Sora platform. So much so that when I see a wonderful video in Moments or Weibo, my first reaction is: Sora did it?
Disney caught off guard
Sora's popularity has not only attracted fans and enthusiasts, but also attracted cooperation from giants in the film and television industry. In December 2025, Disney announced a three-year cooperation agreement with OpenAI - unprecedentedly authorizing more than 200 of its characters to enter the Sora platform, covering Disney, Marvel, Pixar and Star Wars series.
In addition, Disney plans to invest $1 billion in OpenAI. The deal shocked Hollywood. For OpenAI, obtaining Disney IP authorization means that Sora can legally become a creative engine for pop culture; for Disney, this is a proactive embrace of the traditional entertainment empire into the AI era.
However, just 103 days after reaching a cooperation agreement with Disney, Ultraman categorically closed Sora. According to Reuters, just half an hour before Sora announced its closure, the project teams of OpenAI and Disney were still meeting to discuss the details of the cooperation. As a result, Disney was informed half an hour later that the cooperation between the two parties was cancelled.
Although Disney executives were shocked by the sudden halt to the project, their official statement maintained a calm and understanding attitude. "We respect OpenAI's decision to exit the video generation business and pivot to other priorities. We appreciate the constructive collaboration and learnings from both teams during this time. Disney will continue to engage with other AI platforms to find new ways to connect with fans while respecting intellectual property and creators' rights."
In fact, Disney didn't lose much. According to US media reports, Disney’s $1 billion investment in OpenAI is linked to the progress of the Sora cooperation project. The conditions for triggering investment have not yet been met, so OpenAI has not received the $1 billion in funding.
Focus on core and imitate rivals
Why are you so anxious to shut down Sora? The official explanation of OpenAI is: focus on core business and concentrate computing resources. But there is clearer business logic and realistic financial pressure behind this.
At the beginning of 2026, Anthropic's annual revenue has reached US$19 billion, relying on only text and code, no video generation, no image generation, and no consumption of social apps. Claude has only one product interface: chat, code, and computer use, all in one App.
OpenAI looked at where the largest revenue growth in the AI market comes from, and looked at the business model of competitor Anthropic, and the answer was very clear - coding and the enterprise market. So Ultraman began to fully copy this model: merging ChatGPT, Codex and browser into one App.
In addition to Sora, they also turned off the in-app purchase payment feature Instant Checkout. Almost all consumer-side experimental projects are being axed. What is left is the core product that Ultraman wants, which is Anthropic's style of play: an App that focuses on code and chat, focusing on enterprises and developers.
After shutting down Sora, OpenAI's product line became clearer: ChatGPT, o-series inference models, enterprise APIs, and Codex programming tools. This is a company focused on text and code, not a platform that wants to get into social media and entertainment content. Anthropic's Claude family and Google's Gemini are all pressing hard on this core battlefield. OpenAI needs to focus its limited chips and engineers where they can generate real business value.
Video generation is one of the most computationally intensive tasks in AI applications. The computing resources consumed to generate a high-quality video can easily be converted into dozens or even hundreds of GPT-4o conversations. Sora consumes a lot of GPU computing power every month - this computing power could have been used to support ChatGPT's 900 million weekly active users, to train stronger inference models, and to serve enterprise customers. After all the calculations, turning off Sora is a multiple-choice question that is not difficult to choose.
Preliminary loss of US$14 billion this year
Meanwhile, the popularity of video-generated social media didn’t last long. Last month, Sora's monthly downloads have plummeted by more than 65% from the peak, from more than 3.3 million downloads in November 2025 to the current 1.12 million downloads. Monthly active users have also dropped from the peak of 25 million to around 8 million to 10 million.
In fact, OpenAI burns 10 million to 15 million US dollars every day on Sora, and its annual investment is expected to exceed 5 billion US dollars. More than half a year after its launch, the total revenue of the Sora platform is only US$2.1 million. This number is not enough to cover the cost of OpenAI for one day.
The deeper pressure comes from the financial side. OpenAI is burning money at an alarming rate. Last year, their total revenue reached US$20 billion, setting a record for growth, but they lost nearly US$5 billion; and this year's expected loss is as high as US$14 billion, which is equivalent to burning US$47 million every day.
Wall Street has long noticed a paradox: the scale of financing cannot keep up with the speed of cash burning. OpenAI itself projects that total expenditures between 2024 and 2029 may be as high as $488 billion in computing power alone, while expected revenue for the same period is only $345 billion.
OpenAI will complete its latest $10 billion financing next week, with its latest valuation reaching an astonishing $850 billion, bringing the total funding in this round to $120 billion. But in fact, the US$100 billion previously promised by Nvidia, SoftBank and Amazon has not been fully received. As was the case last year, SoftBank’s $30 billion may have to wait a while longer.
At the end of last year, OpenAI’s cash balance was around US$30 billion. Under such circumstances, the latest US$10 billion is more like an emergency payment for OpenAI, mainly coming from venture capital funds such as Andreessen Horowitz, TPG, and MGX. This is enough to support OpenAI’s capital needs in the next few months, and there is time to wait for SoftBank’s investment to arrive.
Preparing for listing requires weight loss
OpenAI has significantly scaled back recently to save valuable money. They gave up Oracle's expansion plan for the Texas data center, started an advertising revenue project that Altman originally disliked, launched the ChatGPT in-app shopping project, and postponed the AI hardware project originally released this year. Closing Sora is only Ultraman's inevitable choice to save money.
Of course, litigation risk is also one of the considerations for closing Sora. During the operation of Sora, the families of Martin Luther King Jr. and Robin Williams publicly posted on social media asking users to stop using AI to generate videos of their late fathers. Complaints of copyright infringement followed. Even if a licensing agreement is signed, the copyright risk of a large amount of user-generated content on the Sora platform is still like a time bomb.
It must be mentioned that OpenAI is very likely to IPO this year. At the critical stage of preparing for listing, shutting down a product that burns huge amounts of money, suffers significant losses, has stagnant growth, and faces the risk of litigation is a rational operation to save money, beautify financial reports, and boost investor confidence.
Let the Sora research team turn to the direction of world simulation and robots, which not only retains the technical narrative, but also gets rid of a commercial baggage that eats resources and causes trouble. From this perspective, closing Sora is not a strategic failure, but a necessary downsizing before the IPO.
Social giants and vertical tools
Sora's sad exit has drawn a clear life-and-death line for the entire AI video industry: there is almost no way out for consumer-level independent applications, and only integration into social giant platforms can continue; only professional tools and B-side integration can ensure a stable revenue market.
Google's Veo series has been deeply integrated into YouTube's creative workflow, and ByteDance's SeaDance has recently shocked Hollywood with a series of professional film and television content with its ultimate camera movement control. In the vertical field, professional tools such as Runway Gen-3 and Kling are entering the professional film and television industry process with more accurate physical simulation and editing functions. Instead of getting colder, the battlefield has become unprecedentedly hot due to a more pragmatic closed-loop business. "
Google's chance of winning lies in ecological integration. Embedding Veo into the YouTube creator tool chain and connecting it with Google Workspace is a sustainable B-side monetization path, rather than relying on an independent application to hit the consumer market head-on. ByteDance has the traffic moat of TikTok and Douyin, and can directly turn AI video tools into content creation productivity, completing a closed loop on its own platform. Both of these paths cannot be replicated by Sora.
Film and television production is the most realistic paid scenario. Major Hollywood production companies spend hundreds of millions of dollars on visual effects every year. If AI video tools can reliably complete pre-concept preview, storyboard production, background replacement, etc., it will be enough to generate considerable enterprise-level subscription revenue. Advertising marketing is another deterministic track. Brands spend huge amounts of money on video advertising content production every year. AI video can significantly reduce the production cycle and cost, and the business logic is clear.
The failure of Sora exposed the fatal dilemma of general consumer AI video products: technical barriers will be quickly leveled, user novelty fades very quickly, the copyright minefield of the content ecosystem is full of pitfalls, and a monetization model is almost impossible to establish. Who will pay for AI videos? Ordinary users are unwilling to continue paying, advertisers are shunned by platforms full of deep fake risks, and content platforms have not yet found a feasible way to integrate AI videos into existing business systems.
Sora completed the entire story of an AI star product in two years: technological shock, capital carnival, user influx, stagnant growth, accumulation of legal risks, absence of business model, and finally quietly offline. This is not the end of the AI video era, but the kind of ambitious narrative that is in the name of subverting everyone but lacks a closed loop of monetization is beginning to accept the calmest judgment of the market.