OpenAI Shuts Down Sora as Disney Exits $1 Billion AI Video Deal

AI

by Nino LeitnerMarch 25^th, 2026

Subscribe to News - Software

OpenAI is discontinuing Sora, its AI video generation platform, consumer app, and developer API, in a dramatic strategic pivot that also torpedoes Disney’s $1 billion investment deal. The company says it will redirect computing resources and the Sora team toward productivity tools, agentic software, and robotics research as it prepares for a potential IPO later this year.

The shutdown marks one of the most significant retreats in the brief but turbulent history of generative AI video. Sora was first previewed in February 2024 as a text-to-video model capable of producing footage with an unprecedented level of realism, and the demos sent shockwaves through the filmmaking community. When the consumer app and Sora 2 model launched in September 2025, complete with a TikTok-style social feed and a “cameo” feature that let users insert their own likeness into AI-generated scenes, it briefly topped the Apple App Store charts. Now, barely six months later, it is being wound down entirely.

Why OpenAI pulled the plug

According to the Wall Street Journal, CEO Sam Altman announced the changes to staff on Tuesday, writing that the company would wind down all products built on its video models. That includes not just the consumer app but also the developer-facing Sora API and video functionality within ChatGPT. The decision is part of a broader strategy shift to redirect computing resources and engineering talent toward productivity and coding tools as the company eyes an initial public offering potentially as soon as Q4 2026.

OpenAI’s applications chief, Fidji Simo, reportedly told employees in an all-hands meeting earlier this month that the company could not afford to be distracted by “side quests.” Last week, OpenAI announced it was merging its ChatGPT desktop app, coding tool Codex, and web browser into a single consolidated “superapp,” and Sora was notably absent from the lineup. In hindsight, the writing was on the wall.

The underlying economics also tell a clear story. OpenAI reportedly pulled in roughly $13 billion in revenue last year but expects to spend approximately $100 billion over the next four years, much of it on data center expansion. Running an AI video generation service, particularly a consumer app with no meaningful revenue stream, is enormously compute-intensive. AI video generation requires far more processing power and electricity than conventional internet services, and Sora’s massive GPU appetite made it increasingly difficult to justify against products with clearer paths to profitability.

The Disney deal collapses

Perhaps the most significant casualty beyond Sora itself is the Disney-OpenAI partnership announced in December 2025. Under that three-year agreement, Disney pledged a $1 billion equity investment in OpenAI, and the two companies planned to license more than 200 characters from across Disney, Marvel, Pixar, and Star Wars for use on the Sora platform. Users were meant to be able to generate AI videos featuring characters from Mickey Mouse to Darth Vader, with curated selections appearing on Disney+. At the time, the deal was widely seen as a watershed moment for both the tech industry and Hollywood.

That deal is now dead. According to The Hollywood Reporter, Disney has confirmed it is exiting the investment. A Disney spokesperson acknowledged the partnership’s end while keeping the door open for future AI collaborations, stating that the company would continue to engage with AI platforms while “responsibly embracing new technologies that respect IP and the rights of creators.”

A turbulent trajectory from the start

Sora’s entire lifecycle was marked by controversy, particularly around copyright and intellectual property. We covered the training data ethics debate extensively when OpenAI’s then-CTO Mira Murati could not confirm whether YouTube videos were used to train the model. When Sora 2 launched last fall, OpenAI initially deployed it without guardrails to protect copyrighted content, immediately sparking a fierce copyright reckoning with Hollywood. Studios, agencies, and unions including SAG-AFTRA, the MPA, CAA, and WME all condemned the platform’s opt-out approach to IP protection. OpenAI eventually added controls allowing copyright holders to block the use of their likenesses and intellectual property, and the Disney deal was supposed to chart a new path forward, but it clearly was not enough to sustain the venture.

The broader AI video landscape has not stood still either. ByteDance’s Seedance 2.0 generated massive controversy of its own in February, with viral deepfake videos of Hollywood stars triggering industry-wide backlash. Google’s Veo remains the most significant player still operating at scale in the AI video generation space, though it has faced its own legal challenges from IP holders.

What happens to the technology

OpenAI says it is not abandoning video generation research entirely. In a statement, the company indicated that the Sora research team will pivot to “world simulation research to advance robotics that will help people solve real-world, physical tasks.” Because video models can serve as reasonable simulations of the physical world, they are increasingly used to train robots for specific tasks. According to Altman, the Sora team will now prioritize longer-term bets such as robotics.

The New York Times also reports that OpenAI will continue to use video generation technologies behind the scenes as a training mechanism for robotic systems, even as the consumer and professional-facing products disappear. What remains unclear is whether the Sora model itself will be preserved, folded into another system, or retired entirely.

What this means for filmmakers

For the filmmaking community, Sora’s demise carries mixed implications. On one hand, it removes one of the most prominent, and for many creators most threatening, AI video tools from the market. On the other hand, the underlying technology is not going away. Google continues to develop Veo, ByteDance’s Seedance family advances despite legal pushback, and numerous smaller players continue to iterate on generative video. The competitive landscape is simply shifting.

For those who want to understand how AI video tools can fit ethically into a filmmaker’s workflow, Directing the Future: Ethical AI Video for Filmmakers by cinematographer Drew Geraci on MZed remains an excellent resource. The course, which is continuously updated with new lessons covering emerging tools, offers practical guidance on using generative video for concept work, storyboarding, and pre-visualization while maintaining ethical boundaries. Drew’s lesson on Sora 2 demonstrated how the tool could be useful for brainstorming while explaining why using it for final content production raised serious ethical concerns.

The speed at which Sora went from industry-shaking debut to shutdown is a potent reminder that the AI video revolution, while inevitable in some form, is not a straight line. Compute costs, copyright battles, and the simple absence of a sustainable business model have claimed what was once the most high-profile name in AI-generated video.

What do you make of Sora’s shutdown? Does it change your outlook on AI video tools for your filmmaking workflow, or is this just a speed bump on the road to widespread adoption? Don’t hesitate to let us know in the comments below!

Featured image: AI-generated with Google’s Gemini “Nano Banana 2”, by CineD (No OpenAI tool was used in the generated of this image.)