As the popularity of AI-powered image and video creation tools rapidly grows, the sheer financial burden behind this technology is also becoming apparent. According to a new analysis, OpenAI’s Sora AI model, which generates compelling videos using text commands, costs the company approximately $15 million per day. This translates to an annual cost of approximately $5.4 billion.
Sora Could Put OpenAI in a Difficult Position
Video-generating models like Sora demand thousands of times more processing power than text-based AIs that simply establish semantic relationships between words. This is because motion, lighting, texture, and physical interactions within a three-dimensional scene are recalculated dozens of times per second in each frame of a video.
Video models must also process time as the fourth dimension; they must not only create a single image but also ensure that a sequence of these images generates a meaningful action. This complex process requires the models to spend significantly more time on the Graphics Processing Unit (GPU), which in turn increases costs.
The most striking aspect is the scale of production. The Sora app has an estimated 4.5 million active users. SemiAnalysis analyst AJ Kourabi notes that approximately a quarter of these users create an average of 10 videos per day.
This translates to over 11 million videos being produced daily. Multiplied by the $1.3 cost of each video, OpenAI spends $15 million daily on Sora alone.
Sora’s lead developer, Bill Peebles, openly admitted in an October statement that the model’s current economics are “completely unsustainable.” However, it’s believed the company deliberately adopted this high-cost strategy.
Mizuho analyst Lloyd Walmsley stated, “The primary goal isn’t cost, but audience building.” He added that OpenAI aims to solidify its market dominance by rapidly growing its user base, then transition to paid subscription plans.
