Ming-Chi Kuo: Nvidia's (NVDA.US) new AI chip, B300, is anticipated to enter mass production in the third quarter.

Earlier, well-known analyst Ming-Chi Kuo published an article to predict the key points of Nvidia's GTC2025. He said that Nvidia's new AI chip B300 is the key focus of the event, including CoWoS-L and CoWoS-S, the biggest highlight is that HBM is upgraded from 192GB to 288GB, and the computing efficiency is improved by 50% compared with B200. B300 is expected to be put into trial production in 2Q25 and mass production in 3Q25.

Moreover, edge AI is also a long-term key trend of Nvidia, but it is expected that GTC2025 will focus on AI servers, and the market expects AI PC solutions N1X and N1 are more likely to be announced at Computex this year.

Kuo said that there are three major concerns about Nvidia's investment trend in AI servers, namely the validity of the scaling law, the mass production of new AI servers, and geopolitics. With the growth of terminal/low-end AI devices, if Nvidia can provide new views on the validity of the scaling law of AI servers, it will help reduce market concerns.

Kuo said that the current mass production of GB200NVL72 is not good, and if Nvidia can provide cases of GB200NVL72 in data centers, focus on the conversion benefits of B200 and B300, and improve the visibility of B300's mass production, it will help the market start to react to the investment theme of B300. Besides, he does not expect Nvidia to address geopolitical issues.

Terminal AI is also a long-term key trend of Nvidia, but Kuo expects GTC2025 to focus on servers, and the market expects AI PC solutions N1X and N1 are more likely to be announced at Computex this year. As AI-related stocks have recently corrected, GTC should provide a catalyst for short-term share price rise, but whether the momentum can be significantly extended depends on whether the content of GTC meeting can reduce investors' concerns above.