Every Daily Hot Comment | Sora Withdraws Quietly, AI Must Ultimately Say Goodbye to Vanity and Return to Business Fundamentals

Daily Economic News Commentator Zhao Linan

On March 24, OpenAI announced that it would shut down Sora and its related APIs—video generation applications that once went viral worldwide.

From the “walking in the streets of Tokyo” demo that stunned the world in early 2024 to now, Sora’s life cycle has lasted only a little over two years. If we count from the end of September 2025 when Sora’s standalone application was released, then less than half a year has passed. Many people feel regret or confusion about this: a product that once represented the pinnacle of human visual generation technology—why has it ended up in such a situation?

Peeling back the surface noise and examining the event from the underlying logic of business, we find that Sora’s exit is like a loud warning bell, signaling that the “flashy, gaudy era” in the AI industry, where people blindly chase spectacle, is coming to an end.

From a strategic perspective, the main battleground of AGI (artificial general intelligence) is the real core, and flashy skills don’t last.

When discussing the reasons for Sora’s failure, we must first clarify a key proposition: where exactly is the moat of a top AI company? Over the past two years, Sora has undoubtedly been the most dazzling “decorative vase” in the tech world. In a very short time, it generated highly realistic videos, greatly satisfying the public’s curiosity.

However, OpenAI’s compute requirements are rapidly expanding, forcing its R&D focus to shift to “world simulation” and deeper, lower-level “robotics technology” and AGI. This is a sober strategic decision—in an age where compute is supreme power, spending extremely valuable GPU (graphics processing units) resources on letting netizens produce funny short videos is a massive waste of a company’s core competitiveness.

Evidently, the main battleground is AGI, and flashy skills can’t last. In OpenAI’s grand narrative of moving toward general artificial intelligence, Sora gradually deviated from the main plot, turning into a bottomless pit that consumes compute. Eliminating this distracting item early is the most responsible strategic retreat for both shareholders and the industry.

From the market logic perspective, AI products need to follow first principles and meet users’ demand for “pixel-level” control over content.

If compute squeezing is the internal reason for Sora’s exit, then its fundamental defect in product philosophy is the direct external reason for why it was discarded. Looking back at the past, Sora has been plagued by controversy. From social panic triggered by deepfake videos to countless low-quality “AI junk” filling the internet, Sora’s loss of control revealed a fatal weakness of today’s generative video models: lack of controllability.

For B-side professional customers (such as advertisers, film production companies, and game developers) who are truly willing to pay for the technology, what they need has never been “blind-box” random generated videos. The film and television industrial system operates under strict rules: directors need actors to make specific expressions at specific frame counts, and lighting designers need light sources to be cast from specific angles.

And Sora’s model is: “input a piece of text and hope it generates the result you want.” If the result is wrong, users can only modify the prompt and try again, but they can’t fine-tune a particular local part of the video the way they would use a scalpel. This lack of a “pixel-level control” black-box mechanism means Sora is destined to stay at the entertainment level of social media, unable to deeply embed into the production workflow of modern industry.

And when you return to the essence of cash flow, AI products must have self-sustaining “cash-generating” capability—which requires companies to bring their attention back to the most fundamental financial statements.

The Sora phenomenon perfectly illustrates what a “trinket/limb” means in today’s AI industry—tasty if eaten? no; regrettable if discarded. In other words: it’s flavorless to use and a pity to throw away.

Maintaining a video generation platform at the level of Sora carries an unbearable reasoning cost behind the scenes. Each time a video is generated, it burns electricity bills and chip depreciation costs. However, its path to monetization is extremely unclear. Faced with high costs, ordinary C-side users have no ability to afford expensive subscription fees; and due to the aforementioned lack of controllability, high-net-worth B-side enterprises also dare not scale it up for commercial use.

The reality Sora faces is that it has become a massive burden that only knows how to burn money, yet cannot generate positive cash flow.

Sora’s exit, in this slightly wintry early-spring March, pours a bucket of ice-cold water into the feverish AI industry. But this is absolutely not an AI winter; it is the “adult coming-of-age ceremony” of the industry moving toward maturity. With farewell to the flashy Sora, the hard-core “great voyage era” of AI pragmatism is only just beginning.

Disclaimer: The content and data in this article are for reference only and do not constitute investment advice. Please verify before using. Proceed at your own risk.

Source of cover image: Economic Daily News

Endless information and precise interpretation—available in the Sina Finance APP