Sora Is Dead. OpenAI Exits AI Video Generation Entirely.
OpenAI shuts down the Sora app, kills the Disney deal, and pivots to robotics. The highest-profile AI video product lasted just six months. Here is what creators need to do right now.

What Just Happened
On March 24, 2026, OpenAI announced it is shutting down the standalone Sora app and its API. The app launched in September 2025 to massive fanfare, shooting to the top of the iOS App Store within hours. Six months later, it is over.
The reason is blunt: compute costs were unsustainable. Generating video at $0.10 per second for 720p adds up fast when millions of users are creating content. OpenAI confirmed the Sora research team will now focus on "world simulation research to advance robotics," a pivot from consumer video creation to infrastructure research.
The Disney Deal Collapse
Perhaps the bigger story is the unraveling of the Disney partnership. Announced in December 2025 with a reported $1 billion investment, the deal would have let users generate content featuring Disney's 200+ iconic characters through Sora. It was the most ambitious content licensing deal in generative AI history.
That deal is not moving forward. Reports from Bloomberg and Variety confirm both sides have stepped back. Disney invested heavily in character safety pipelines and brand protection layers. OpenAI could not guarantee the compute economics would work at scale.
This is not just a product shutdown. It is a signal that the standalone AI video app model faces serious challenges. Building a dedicated platform for AI video generation requires massive GPU infrastructure with thin margins. Even OpenAI, with its billions in funding, could not make the numbers work.
Why Not ChatGPT?
You might expect the DALL-E playbook here: shut down the standalone app, fold video generation into ChatGPT, reach 900 million weekly active users. That is not what is happening.
According to NBC News, ChatGPT will also no longer generate video based on text prompts. Earlier reports in March suggested OpenAI was planning a ChatGPT integration, but the March 24 announcement made clear the company is exiting video generation, not relocating it.
The reason comes back to cost. DALL-E generates static images, which are computationally cheap compared to video. A single video clip requires orders of magnitude more GPU time than an image. Offering video generation to hundreds of millions of ChatGPT users at the current cost structure would be financially reckless, even for a company that just raised $110 billion.
What This Means for Your Workflow
Here are the practical implications.
If You Were a Sora User
Your existing workflows need to move. The API is shutting down, which means any production pipelines built on Sora need alternatives immediately. The good news: you have options that did not exist six months ago.
If You Use Other Tools
The competitive field just narrowed. One fewer major player means the remaining tools, Runway, Kling, Veo, and Adobe Firefly, face less pressure on pricing but more scrutiny on quality. Expect each to position itself as the "Sora alternative" in marketing.
If You Are Building a Product
The standalone AI video app model just took a serious hit. Sora was the highest-profile attempt, backed by the most well-funded AI company on the planet, and it could not sustain the economics. This does not mean AI video products are doomed. It means the future looks more like:
The Bigger Picture: Three Trends Colliding
The Sora shutdown did not happen in isolation. Three forces are reshaping AI video right now.
1. Consolidation into mega-platforms
Adobe Firefly now hosts 30+ models including Runway Gen-4.5, Kling 2.5, and Google Veo 3.1. Google Flow merged Whisk, ImageFX, and Veo into one workspace. The era of standalone single-purpose AI video tools is ending. Sora's shutdown is the clearest proof yet.
2. Real-time generation changes the game
At GTC 2026 last week, Runway demonstrated real-time video generation on NVIDIA's Vera Rubin GPUs, achieving time-to-first-frame under 100 milliseconds for HD video. When generation becomes instant, the entire user experience shifts from "submit a prompt and wait" to "direct a scene in real time." This makes dedicated generation apps less necessary.
3. The cost problem is not solved
Sora's shutdown confirms what the industry has been whispering: AI video generation is expensive. Really expensive. The models that produce the best results consume enormous GPU resources. Until inference costs drop significantly or more efficient architectures arrive, the economics will favor platforms that can spread video generation costs across many other revenue-generating features.
Where Should Creators Go?
If you need a new home for your AI video work, here are your options:
| Tool | Best For | Price Range | Key Strength |
|---|---|---|---|
| Runway Gen-4.5 | Professional quality, real-time future | $12-76/mo | Cinematic quality, NVIDIA partnership |
| Google Veo 3.1 | YouTube creators, integrated workflow | Free-$20/mo | YouTube integration, native audio |
| Kling 3.0 | Character consistency, multi-shot | $5-50/mo | 4K at 60fps, storyboarding |
| Adobe Firefly | Multi-model access, existing Adobe users | $10+/mo | 30+ models, custom training |
| Bonega | Unlimited duration, Veo 3 access | Flexible | No duration limits, multi-model |
My Take
The Sora shutdown feels like the end of a chapter, not the end of a story. The standalone "AI video generator" app was always an awkward product category. Video generation is a capability, not a destination. It belongs inside the tools where people already create, edit, and share content.
What surprises me is how complete the exit is. No ChatGPT integration, no scaled-down version, no API maintenance mode. OpenAI looked at the numbers and decided AI video is not worth the compute at any scale. That is a cold calculation from the company that hyped Sora harder than anyone.
The real question is what happens next. If the best-funded AI lab in history cannot make video generation profitable, who can? Runway is betting on real-time generation with NVIDIA's hardware. Adobe is betting on being the platform layer that aggregates everyone else's models. Google is betting on YouTube integration to justify the cost through ad revenue.
One of those bets will pay off. But Sora's six-month run proves that raw model quality alone is not enough. The business model matters just as much as the model weights.
Sources
- Bloomberg: OpenAI moved to discontinue Sora six months after launching its standalone app and wound down… (Bloomberg)
- OpenAI: OpenAI priced Sora 2 API video generation at $0.10 per second for 720p portrait or landscape… (OpenAI)
- The Walt Disney Company: Disney's original Sora agreement involved a $1 billion OpenAI investment and more than 200… (The Walt Disney Company)

Creative technologist from Lausanne exploring where AI meets art. Experiments with generative models between electronic music sessions.
View profile →Related Articles
Continue exploring with these related posts

10 Best Sora Alternatives in 2026, Ranked (Updated July)
Sora shut down March 24, 2026. We ranked the 10 best replacements — Veo 3.1, Runway, Kling 3.0, Bonega, Luma — by output quality, real pricing, and duration limits, with a quick pick per use case.

Bonega vs Sora (OpenAI) in 2026: The Complete Comparison for AI Video Creators
How does Bonega.ai compare to OpenAI's Sora for AI video generation? We compare pricing, features, availability, and output quality to help you choose.

Meta Mango: Inside the Secretive AI Video Model Aiming to Dethrone OpenAI and Google
Meta reveals Mango, a new AI video and image model targeting 2026 release. With Scale AI co-founder Alexandr Wang at the helm, can Meta finally catch up in the generative AI race?