Why OpenAI Shut Down Sora After 6 Months and Disney’s $1B Deal Crumbled

1. Announcement and Immediate Impact

On March 24, 2026 OpenAI posted a farewell message on X, stating that the Sora app, its website, and API will be fully discontinued, with no specific timeline for data‑export tools. The same day Disney announced it was ending its partnership with OpenAI, effectively cancelling a previously announced $1 billion investment tied to Sora.

2. Stated Reason: Compute Shortage

OpenAI’s official comment cited growing compute demands and a strategic shift toward world‑simulation research for robotics, implying that Sora’s video generation was too compute‑intensive to continue alongside next‑generation language model development.

Interpretation: Sora consumed a disproportionate amount of GPU resources that could be redirected to training the upcoming large model codenamed “Spud.”

According to the Wall Street Journal , internal staff had long complained that Sora was “dragging down the company’s compute resources,” and Sora lead Bill Peebles had previously announced limits on video generation due to chip shortages.

3. Disney’s $1 B Investment Fallout

In December 2025 Disney publicly pledged $1 billion to collaborate with OpenAI, allowing the use of Disney characters to enrich Sora’s content ecosystem. The deal never materialised; no funds were transferred and the partnership was terminated alongside Sora’s shutdown.

Disney’s statement emphasized a commitment to “responsibly embracing new technologies that respect IP and the rights of creators.” The article links this to a deeper concern: Sora’s deep‑fake capabilities conflicted with Disney’s IP‑centric business model.

4. Deepfake Controversy

Sora 2’s technology was praised for its quality, yet it enabled users to generate realistic videos of deceased or copyrighted personalities (e.g., Michael Jackson, Martin Luther King, Pikachu). This sparked protests from families, performers’ unions, and rights holders.

OpenAI later restricted public‑figure video generation, but the action came after a public outcry, leading critics to label it a “post‑hoc fix.”

5. User Retention and Product‑Market Fit

The app’s adoption curve resembled a roller‑coaster: 1 million downloads within five days, decent first‑day retention, but a steep drop to roughly 1 % after 30 days. Users primarily created novelty videos for social sharing, a low‑frequency use case compared to ChatGPT’s daily conversational tasks.

Technical shortcomings—such as distorted faces, missing objects, and broken physics—further eroded user satisfaction, turning impressive demos into disappointing experiences.

6. The Broader AI Video Landscape

Despite Sora’s shutdown, the AI video market remains active. Competitors include Google Veo 3.1 (integrated with YouTube), Runway Gen‑4 (favoured by creators), Kuaishou’s Ke‑Ling AI (dominant in China), and ByteDance’s “Dream” project. All face the same fundamental question: can flashy video generation translate into a profitable business?

Analysts suggest that AI video may re‑emerge as a component of larger “super‑apps” like ChatGPT rather than as a standalone product.

7. User Content Concerns

The shutdown notice warned that OpenAI is “exploring” ways to let users export their creations, but no concrete solution has been offered. Men’s Journal warned that creators could lose all their work, urging manual backups.

Community member Ryan T. Brown speculated that a massive lawsuit might be driving the rapid shutdown.

8. Editorial Takeaways

The article summarises four key lessons:

Demo ≠ Product: Stunning technical demos do not guarantee sustainable commercial success.

Compute Is a Hard Constraint: Limited GPU resources force companies to prioritise language and code models over video.

IP and Deepfake Are Minefields: Traditional media owners view AI‑generated content as a high‑risk infringement.

IPO Pressures Reshape Priorities: In a pre‑IPO phase, cash‑burning projects without clear returns are often cut.

Ultimately, AI video technology is unlikely to disappear; it simply has not yet found the right product form or business model, much like early PDAs before the smartphone era.

9. Timeline of Sora

2024‑02: Technical preview released, generating 60‑second videos from text.

2025‑09: Sora 2 launched; iOS app, API, and web interface go live.

2025‑09 (5 days later): Over 1 million downloads, topping the App Store free chart.

2025‑10: Deepfake controversy erupts; public‑figure video generation limited.

2025‑12: Disney announces $1 billion investment tied to Sora.

2026‑03‑24: OpenAI announces Sora shutdown; Disney ends cooperation.