The Shutdown of Sora

On March 24, 2026, OpenAI announced the unexpected shutdown of its AI video generation platform, Sora, just six months after its launch. In a statement, OpenAI expressed gratitude to users who created and shared content on Sora, while also indicating that details regarding the application and API shutdown would be shared soon, along with plans for preserving user-generated content.

CEO Sam Altman clarified in an internal meeting that the shutdown would affect both consumer and developer versions of Sora, including the video functionality within ChatGPT. This decision indicates a significant shift in OpenAI’s focus, as the video generation product line is being removed from its core offerings. However, the technical team remains intact, redirecting efforts towards foundational research in robotics and world modeling.

The announcement took many by surprise, with reports suggesting that several Sora team members were only informed of the closure on the morning of the announcement. Just days prior, Sora was still actively engaging with content partners, including Disney.

Sora’s Development Journey

Sora’s trajectory seemed promising. In February 2024, OpenAI released the Sora 1.0 technical preview, which utilized an innovative Diffusion Transformer architecture capable of generating detailed video clips up to 60 seconds long. Following over a year of iterations, Sora 2 launched in September 2025 as a standalone social application, introducing advanced features such as audio-visual synchronization and real human character integration, quickly rising to the top of the Apple App Store’s free chart.

In December 2025, OpenAI secured a three-year partnership with Disney, which included a $1 billion investment and access to over 200 classic IPs from Disney, Marvel, Pixar, and Star Wars for content generation on the Sora platform. This collaboration was seen as a pivotal moment for AI video becoming mainstream.

However, this vision abruptly ended with Sora’s closure, and Disney acknowledged OpenAI’s decision to shift its business priorities and exit the video generation space.

An AIGC director expressed regret, stating, “If it had continued, Sora could have become very strong, but it chose not to move forward.”

The Decline Before the Official Shutdown

Before OpenAI’s official announcement, Sora’s decline was already evident within the AIGC video creation community. Many AIGC directors reported a noticeable drop in Sora’s usage over the past few months, with it gradually being phased out of daily production workflows.

The primary issue was clarity. Sora’s generated videos struggled with detail reproduction, texture stability, and overall cleanliness, often resulting in blurry faces and breakdowns in complex scenes. This indicated that Sora functioned more as a concept validation tool rather than a reliable production tool.

The more critical factor was that competitors had surpassed it. AIGC director Ding Yi’s experience is illustrative. Initially impressed by Sora 2’s capabilities, he soon transitioned to using Seedance, especially after the release of Seedance 2.0, which offered higher resolution outputs, more stable long-shot control, and better semantic understanding.

Sora’s initial market excitement quickly faded. Despite a strong launch, Appfigures reported a 32% drop in downloads by December 2025, followed by a further 45% decline in January 2026. Even with Disney IP integration, Sora could not regain its footing.

The Cost of Sora’s Ambitions

OpenAI reportedly spent between $10 million and $15 million daily on Sora, amounting to approximately $5.4 billion annually. This investment in a project with unclear returns raised questions about its sustainability.

When a product fails to establish irreplaceability in production and does not create a retention loop in consumption, its fate may already be sealed.

Strategic Retreat for OpenAI

The closure of Sora reflects a strategic contraction for OpenAI, which is not an isolated event. On March 19, OpenAI announced the integration of its desktop ChatGPT, Codex, and Atlas browser into a single “super app.” Product lead Fidji Simo noted that the previous fragmented product lines were slowing overall progress and affecting product quality.

In a broader context, OpenAI’s recent moves indicate a shift towards focusing on core competencies and maximizing efficiency, particularly with the anticipated release of its next-generation model, codenamed “Spud.”

The Aftermath: Diverging Perspectives on Video AI

Sora’s shutdown has sent ripples through the AIGC video sector, prompting varied reactions. Some see an opportunity for domestic video model players to capture market share, while others remain skeptical about the viability of video generation as a sustainable business model.

The competitive landscape for video models is challenging, requiring continuous investment to maintain a leading position. As one industry expert noted, the competition is fierce, and without adequate funding or rapid research progress, any lagging aspect can lead to a swift loss of users to more advanced models.

In summary, the post-Sora era reveals a dual reality: while technological advancements continue, the path to commercialization remains fraught with difficulties. The balance between innovation and sustainable growth will determine the future of video AI.