In the AI video generation arms race, OpenAI is burning money at an astonishing speed to seize the market. According to a recent estimate by Forbes, the company's hit application Sora has been costing $15 million per day since its launch on September 30, with annual expenses potentially exceeding $5 billion. Although the app has surpassed 4 million downloads in just one month and generates millions of 10-second short videos daily, this traffic frenzy is backed by an unsustainable financial black hole and imminent strategic adjustments.

The "Fictitious Prosperity" Behind the Hit

Sora's viral spread has generated a large amount of entertainment content: absurd short films like "cats and dogs having meetings" and controversial videos featuring "resurrected deceased stars," with users showing great enthusiasm. However, OpenAI insiders admit that the majority of free users only use it to create funny content, with no intention to pay and difficulty in converting into high-value commercial scenarios. As CEO Sam Altman said, "This content cannot support the computing costs we burn every second."

sora

The Unsustainable "Internet-Style Expansion"

Sora project lead Bill Pibbles admitted that the current "free + high computing power" model is completely unsustainable. Although Bank of America Securities analyst Lloyd Walmsley compared it to the successful early internet strategy of "capturing users first, then finding monetization," the marginal cost of AI video is much higher than text-based content—generating a single high-definition video consumes several times more GPU resources than text, and the curve for unit cost reduction is far flatter than expected.

Strategic Retrenchment Is Inevitable

To control losses, OpenAI has decided to significantly reduce the video generation quota for free users and shift resources toward high-value users such as enterprise clients and film studios. At the same time, the company is exploring a tiered pricing model: basic entertainment content is free, while advanced features such as professional long videos, 4K output, and commercially usable copyrights will be charged at a premium. Additionally, user-generated video data will also be used to improve model training, forming a "use-optimize-monetize" loop.

The Critical Point of AI Video Has Not Yet Arrived

Although Sora is technologically ahead of the industry, the market is not yet ready to pay for it. Current AI video still faces bottlenecks such as length limitations, physical logic flaws, and copyright risks, making it difficult to replace professional production. OpenAI's gamble is essentially betting that within the next 12–18 months, technology, demand, and business models will all mature in sync.

AIbase believes that Sora's dilemma reveals a deep contradiction in generative AI: technological breakthroughs do not equate to commercial success. When the industry moves from "being able to generate" to "being worth paying for," whoever can build a sustainable unit economic model first will truly win the final battle in AI video. Until then, OpenAI must find that dangerous balance between burning money and survival.