Why did OpenAI have to cut Sora?

If someone had told me a few years ago that OpenAI would actively shut down its most “popular” consumer product, I would probably think it was a joke.

After all, when Sora was released at the end of 2024, the sense of shock was real. Those videos of “inserting yourself into movie scenes” went viral on social media faster than any product launch. Within less than five days, downloads exceeded 1 million.

Some call it the “iPhone moment of the AI era.”

However, on March 24th, U.S. time, OpenAI announced it would shut down the consumer app of Sora, the developer version, and the embedded video feature in ChatGPT.

Those who once cheered for Sora are probably now asking the same question:

Why is OpenAI cutting Sora?

01 From “Million Downloads” to “Side Quest”

Sora’s decline actually happened much faster than most people realized.

By December 2025, app downloads had dropped 32%. By January 2026, installs continued to fall by 45%. The absolute number was still around 1.2 million, but the trend told the story:

Users stayed, but enthusiasm did not.

The reason isn’t complicated. The “insert yourself into movie scenes” feature was essentially a great demo, but not a repeatable use case. Most people tried it once, found it amazing, and then didn’t know what else to do with it.

A more practical issue is that many users simply aren’t willing to give their portraits to an AI app for processing. This isn’t a technical problem; it’s a trust issue. Without celebrity faces or IP backing, the content Sora could generate quickly hit a ceiling of “interesting but useless.”

Those content licensing partnerships once hoped for are also quietly loosening.

At the same time, Variety and Hollywood Reporter both reported that Disney announced the termination of its partnership with OpenAI, along with the cancellation of a previously discussed $10 billion investment plan.

Originally, Sora was supposed to generate videos from over 200 Disney, Marvel, Pixar, and Star Wars characters, and earlier this year, it was to offer “fan inspiration” creation through ChatGPT.

It’s said that a Disney team was working alongside Sora’s engineers the night before, only to wake up the next morning to news of the partnership’s end. The suddenness probably affected users even more.

02 Full Throttle Toward IPO

If user attrition is the surface reason for shutting down Sora, then the IPO is the real behind-the-scenes driver.

OpenAI just completed a $11 billion funding round, pushing its valuation to $730 billion. At this point, every bit of resource allocation is scrutinized by investors and potential public market shareholders.

Sora is an extremely resource-intensive product. Generating videos costs much more than generating text, and much more than generating images. Every time a user “plays” with an AI video in Sora, the GPU power burned could be enough for ChatGPT to answer dozens of questions.

When OpenAI’s head of applications, Fidji Simo, clearly stated at a company-wide meeting that the company needs to stop being distracted by “side tasks” and aggressively shift focus to coding and enterprise users—Sora’s fate was essentially sealed.

It’s not a difficult decision; it’s a clear-headed one.

Diverting compute resources from Sora back to Codex (OpenAI’s AI coding assistant) makes perfect business sense.

Since this year, Codex has tripled in users, grown fivefold in usage, and now has over 2 million weekly active users. Last month, OpenAI also acquired developer tools company Astral, integrating the entire team into the Codex group.

The direction is set.

On the other side, Anthropic provided a more straightforward data point: annual revenue exceeds $19 billion, with about 80% coming from enterprise clients. CEO Dario Amodei mentioned that in February alone, they added $6 billion in revenue, almost all from Claude Code.

Enterprise clients pay, renew, and expand usage. Consumer users may like, share, but not necessarily pay. This has been one of the most painful and important lessons in the AI industry during 2025-2026.

03 Is the “AI Video” Track Cooling Off?

The answer might be no. Or at least, OpenAI has chosen not to wade into that muddy water.

In fact, competition in AI video generation continues and is intensifying. ByteDance’s Seedance 2.0 is still running, and Google DeepMind’s Veo 3 is recruiting filmmakers.

Sora’s exit from the consumer market doesn’t mean this direction is inherently unviable.

But a structural problem facing Sora is one that competitors might not be able to avoid: the issue of “AI garbage.”

When a tool can easily produce hyper-realistic videos, low-quality AI content begins flooding social media. This isn’t just a platform moderation challenge; it’s quietly eroding overall trust in AI-generated content. When “AI creation” becomes synonymous with poor-quality content, a brand’s value for an AI video app can be severely damaged.

Paul Roetzer, founder of Marketing AI Institute, put it well: he believes Sora’s underlying video generation technology is “incredible,” but OpenAI chose to turn it into an endless entertainment scrolling tool—completely opposite to the direction labs should be heading.

This somewhat explains why a technically leading product can still fail commercially.

Sora isn’t lacking in technology; it’s lacking an answer to “why would users keep using it?”

In entertainment and consumer scenarios, that answer remains vague; but in professional creation, film industry, and advertising, the answer can be much clearer.

However, OpenAI now seems unwilling to find that answer itself—at least not in the form of a consumer product like Sora.

Looking back, OpenAI’s product lineup over the past year has been somewhat scattered.

ChatGPT, Sora, image generation, voice assistants, enterprise APIs, Codex, customized GPT… each direction was pursued, each achieved a “decent” result, but “decent” is increasingly less valuable in this industry.

Now, the story has changed.

The shutdown of Sora is a signal and a choice.

OpenAI is telling everyone: we know what makes money, and we’re focusing resources there. Meanwhile, the staff plan to grow from 4,500 to 8,000 by the end of the year, which shows the company isn’t shrinking but concentrating on more important things.

This company is using action to answer a question many AI companies are still avoiding: are you building a tech showcase, or a real business?

Sora’s fireworks were beautiful.

But once they fade, what remains is what the company truly values.