The day before yesterday, I timely reminded about China's AI model DeepSeek potentially impacting Nvidia and the US stock market, and now it has caused a huge effect. I just saw Jiang Zhuoer’s statement that "DeepSeek is beneficial to Nvidia". After checking some information, I would like to refute this: Jiang Zhuoer’s point that "DeepSeek only optimized computing costs and did not achieve any new breakthroughs" is incorrect. In fact, DeepSeek excels not only in significantly reducing the cost of training models and enhancing inference capabilities but also in the fact that traditional models (ChatGPT-4) require dozens of H100 level GPUs (over a million dollars in costs) for full deployment. DeepSeek has broken through the scale limitations of large model deployment through technological innovation, with R1 able to run smoothly on a single RTX 4090 (24GB memory), with an inference speed of 15-20 tokens/second. This is quite frightening for Nvidia. A large portion of their current profits comes from high-performance GPUs, and if everyone can do AI using ordinary gaming GPUs, that would be a big problem. In any case, the capital market has provided the answer~ it has chosen to be bearish on Nvidia. I suggest that we should still observe more at this moment and let the bullets fly for a while.
Disclaimer: Includes third-party opinions. No financial advice. May include sponsored content.See T&Cs.