Trending stories say OpenAI's viral video app Sora costs the company roughly $15 million daily to operate, but is that figure accurate? And other tough questions about AI priorities and resource allocation.
OpenAI's Sora video generation app has become a sensation since its September debut, racking up 4 million downloads by Halloween and churning out millions of 10-second AI videos daily. But behind the viral memes and fantastical content lies a sobering financial reality: News sources are claiming that the company is spending approximately $15 million per day, or $5.4 billion annually, to power the service.
That figure comes from back-of-napkin calculations by analysts who estimate each 10-second video costs OpenAI around $1.30 to generate, accounting for GPU rental, processing time, and computational overhead. For a company that lost more than $12 billion last quarter while projecting $20 billion in annual recurring revenue, the math is stark.
OpenAI's own head of Sora, Bill Peebles, acknowledged the unsustainable economics in late October, yet the company continues offering free video generation to millions of users. The strategy mirrors a familiar Silicon Valley playbook: build audience and engagement first, worry about profitability later. On X, Sam Altman discusses the high levels of investment in technology infrastructure. He also admits that it could be a mistake, saying, "This is the bet we are making, and given our vantage point, we feel good about it. But we of course could be wrong, and the market—not the government—will deal with it if we are."
The hope is that as video models become more efficient, costs will plummet exponentially, and OpenAI can eventually monetize through subscriptions, premium features, or advertising. CEO Sam Altman has already signaled that free access won't last forever, noting that casual users making funny memes for friends cannot be supported by any realistic ad model.
The community response has been skeptical. Commenters have expressed concern about channeling finite energy resources into generating what many describe as frivolous AI content. I've tested it myself, and it's hard to argue that meme videos, often riffs of other videos, have much value beyond the audience on Sora 2 at the moment.
The broader sentiment reflects disappointment that such enormous computational power and investment are directed toward entertainment rather than addressing pressing real-world problems. Observers question whether this represents a misallocation of talent and resources, especially when AI could theoretically be applied to medicine, climate science, or infrastructure challenges. Experts suggest the current spending spree is less reckless than it appears.
Analysts at firms like Mizuho and Cantor Fitzgerald note that video model efficiency could improve fivefold within a year and dramatically more by 2027. Additionally, the free videos allow OpenAI to harvest user data and video descriptions for training its models, potentially providing long-term competitive advantages. The company can also deduct compute costs from its taxable income as it transitions to for-profit status.
Still, the fundamental tension remains: OpenAI is betting billions that future efficiency gains and monetization strategies will justify today's staggering burn rate. Michael Burry bet against the AI bubble, but his public concerns highlighted Oracle and Meta over Microsoft and OpenAI. But the question remains about the true cost of the video generation. The company's 10-Q doesn't break out the cost of Sora alone. And without understanding whether OpenAI has included any profit in their credit system, or the real volume of clip generation, the jury is still out on the actual cost.
The Sora story ultimately reflects a deeper tension in AI development. The technology is undeniably impressive, but questions linger about whether the industry's current trajectory serves society's most pressing needs or simply chases engagement metrics and market share. As OpenAI races to scale and monetize, the community is watching closely to see whether this expensive experiment yields genuine value or becomes a cautionary tale about misaligned incentives in the AI era.
