OpenAI Plans Sora Revival via Hollywood-Backed Startup Led by Original Research Head

OpenAI Considers Reviving Sora via Hollywood-Backed Startup

According to an exclusive report from The Information , OpenAI is discussing a plan that could bring its shuttered video generation model Sora back to life. The potential deal involves a newly formed startup co-founded by former Sora research lead Bill Peebles, Hollywood producer and DreamWorks Animation co-founder Jeffrey Katzenberg, and former Dropbox CFO Sujay Jaswa. The trio's company aims to develop and sell AI video software and services tailored for filmmakers, leveraging Sora's underlying technology. If an agreement is reached, OpenAI would receive an equity stake in the new venture.

Team Combines AI Research, Hollywood Expertise, and Business Operations

The startup's founding team brings together three distinct backgrounds. Bill Peebles led Sora's research at OpenAI, giving him intimate knowledge of the model's architecture and training process. Jeffrey Katzenberg, a former Disney executive and DreamWorks co-founder, contributes deep entertainment industry experience and connections. Sujay Jaswa, previously CFO at Dropbox, adds operational and financial expertise; he and Katzenberg previously co-founded the investment firm WndrCo. This combination positions the company to translate video generation research into professional filmmaking tools.

Bill Peebles

Jeffrey Katzenberg

Uncertainty Around Model Weight Access

A key question in the negotiations is whether OpenAI will grant the startup access to Sora's model weights — the learned parameters that determine model behavior. Model weights are typically a closely guarded core asset for AI companies. The report states that the parties have not yet decided if weights will be shared or if settings that significantly influence model behavior will be opened. However, sources indicate that any deal would provide the startup with "sufficiently deep access" to extensively customize Sora, going far beyond standard API calls. This deep access is feasible because Peebles' prior involvement means the team can build on existing technical knowledge.

Sora's Rapid Timeline: Launch, Iteration, and Shutdown

February 2024: OpenAI first demonstrates Sora, generating high-quality video from text prompts.

December 2024: Sora is officially released to users.

2025: OpenAI launches Sora 2 and builds a standalone app for AI video creation and sharing, targeting broader consumer adoption.

March 2025 (reported): OpenAI announces the closure of Sora's business.

April 26, 2026: Sora web and app services stop.

September 24, 2026: Sora APIs are shut down.

The decision to shut down was driven primarily by cost. Sora requires massive compute resources for both development and user inference. Continuing to operate Sora would have demanded sustained, expensive investment that could otherwise be directed toward ChatGPT, coding tools, and enterprise AI products. As competitors like Anthropic advanced in coding and enterprise markets, OpenAI chose to concentrate resources on higher-certainty businesses.

Fallout from Shutdown: Disney Partnership and $1 Billion Investment Lost

According to a prior Reuters report, Sora's closure also terminated a planned collaboration with Disney that included a prospective $1 billion investment from Disney into OpenAI. Because the partnership had not been finalized, the investment never materialized.

Strategic Rationale for a Spinout

The proposed arrangement offers benefits for both sides. OpenAI can recoup potential returns on its substantial R&D investment in Sora through equity in the startup, while keeping its core team focused on large language models, ChatGPT, and enterprise products. For the new company, deep access to a frontier video model eliminates the need to train a comparable model from scratch, reducing upfront capital and time requirements. However, actual returns depend on the final terms, ongoing R&D costs, and commercial traction.

Pivot from Consumer Video to Professional Filmmaking

If Sora returns, it will likely not resume its previous consumer-facing form. The new team is targeting the professional film production market, which has more specific demands than casual short-video generation. Beyond generating visually appealing clips, a production tool must integrate into real workflows, support iterative revisions, and enable team collaboration. Peebles' technical expertise and Katzenberg's industry knowledge provide a strong foundation for addressing these challenges, but the article notes that no public information yet confirms the quality, control precision, or commercial viability of the planned product.

Competitive Landscape and OpenAI's Broader Strategy

Meanwhile, competitors such as Google continue to advance their own video models, meaning a revived Sora would face a changed market. Notably, OpenAI has recently begun re-expanding into areas it previously scaled back, including robotics and creator-focused products. This suggests that while OpenAI's primary focus remains on ChatGPT and enterprise AI, it is willing to explore alternative structures — such as equity-backed spinouts — to extract value from non-core technologies. The Sora case could become a notable example of whether a frontier model, shelved due to cost and priority shifts, can find a second life in a vertical professional market.

Original source: https://www.theinformation.com/newsletters/ai-agenda/openais-plan-revive-sora