OpenAI’s first public demonstration of text-to-video generator Sora sparked excitement in the crypto market and boosted AI token prices, CoinDesk reported. However, for this technology to become mainstream, the computing power required would be staggering. It is estimated that more server-grade H100 GPUs will be needed than Nvidia produces in a year or that its largest customers run in their data centers. According to CoinGecko data, the AI token category now has a market capitalization of $25 billion. Behind the promise of AI-generated video are massive graphics processing units (GPUs), processors from companies like Nvidia and AMD that make the AI revolution possible because they are capable of computing huge amounts of data. A new research report from Factorial Funds estimates that 720,000 high-end Nvidia H100 GPUs are needed to support TikTok and YouTube's creator communities. Sora requires up to 10,500 powerful GPUs for a month of training, and each GPU can only generate about 5 minutes of video per hour. As more people and companies start using AI models like Sora to generate videos, the computing power (inference) required to create new videos will exceed what was needed to initially train the AI model.
