OpenAI's latest video generation model, Sora, has attracted widespread attention for its powerful generation capabilities. According to Factorial Fund estimates, Sora's high performance means huge computing resource requirements. The article analyzes Sora's computing resource consumption and potential market impact, and predicts future development trends. Let’s take a deeper look at the technology and business model behind Sora.
According to Factorial Fund estimates, OpenAI's Sora can generate 5 minutes of video per hour on NVIDIA H100, which is equivalent to 120 minutes of video per day. To support the creator communities of TikTok and YouTube, it is estimated that approximately 720,000 NVIDIA H100 GPUs will be needed. Sora will be aimed at visual artists and filmmakers, with plans to add sound and editing flexibility. OpenAI is promoting Sora to Hollywood, where creators can generate multiple candidate videos before uploading them.Sora's high performance and wide application prospects make it a leader in the field of AI video generation, but its huge computing resource requirements also imply high costs and technical challenges. How Sora's development direction will evolve in the future deserves continued attention.