Crypto and AI’s High GPU Demand Challenges Tech Giants
The Bottom Line
The world of cryptocurrencies and artificial intelligence (AI) is facing a massive challenge: To achieve the ambitious goals of crypto-AI projects, especially in the field of text-to-video generation, a vast amount of computing power is needed. Current estimates suggest that the required graphics processing units (GPUs) could exceed the entire market value of the AI-token category.
The Challenge: An Army of GPUs
The fascination with generating videos from text has greatly increased following the introduction of a demo by OpenAI. However, to make this technology accessible to the masses, an impressive amount of computing power is necessary. It is estimated that hundreds of thousands of server-grade H100 GPUs are needed – more than large technology companies like Microsoft, Meta, and Google currently use combined.
A report from Factorial Funds highlights that just to support the creative communities of TikTok and YouTube, about 720,000 Nvidia H100 GPUs would be required. For comparison: Nvidia delivered a total of 550,000 of these GPUs in 2023.
The Financial Gap
The cost of acquiring the necessary GPUs is astronomical. With an assumed price of $30,000 per card, it would cost $21.6 billion to realize the vision of AI-generated text-to-video content in the mainstream. This amount is almost equal to the entire market value of the AI-token category, which according to CoinGecko is $25 billion.
Alternatives and Challenges
While Nvidia plays a key role in the AI revolution, it is not the only player in the market. AMD offers competing products, and there are opportunities to outsource computing power from GPU farms, as Render (RNDR) and Akash Network (AKT) demonstrate. However, most GPUs in these networks are gaming GPUs, which are less powerful than Nvidia’s server-grade H100 or AMD’s competing products.
Our Assessment
The vision of AI-generated videos from text is undoubtedly fascinating and has the potential to revolutionize creative workflows in Hollywood and beyond. Nevertheless, the crypto-AI industry faces a massive challenge in terms of the required hardware. It is unlikely that this technology will become mainstream in the near future, as the required computing power and associated costs are enormous. The industry must find innovative solutions to overcome these hurdles.
You should read that too:
Sources: Factorial Funds, Statista