NVIDIA’s CEO Jensen Huang has addressed concerns about the company’s stock performance following the emergence of DeepSeek, an open-source reasoning model. Despite the initial 17% drop in stock price on January 27th, NVIDIA’s shares have since recovered most of their losses.
Huang explained that investors’ perception of DeepSeek was misguided, citing a “mental model” that assumed pre-training and then inference were the only parts of AI development. He claimed this paradigm is incorrect, stating that pre-training and post-training are interdependent and equally important.
The CEO emphasized that post-training is where problems are solved and that the more reasoning done before answering a question, the better it becomes. Huang also highlighted the third scaling law, which suggests that increased reasoning capabilities lead to improved performance.
Huang’s comments came after the initial market reaction to DeepSeek’s R1 model was overly negative, with some predicting the end of AI as we know it. However, the CEO asserts that his company is well-positioned to benefit from this new paradigm, dispelling concerns about its undervaluation.
Source: https://www.investing.com/news/stock-market-news/nvidia-ceo-jensen-huang-explains-how-the-stock-market-got-it-wrong-on-deepseek-3884179