OpenAI's deployment of the Sora model requires huge investment

According to a report by market research firm Factorial Funds, the text-to-video model Sora deployed by OpenAI required 720,000 Nvidia H100 AI accelerator cards at its peak. If each card costs $30,000, the total cost is about $21.6 billion. In addition, running this model requires a lot of electricity, and the power consumption of 720,000 H100 accelerator cards is as high as 504,000,000W.