Workflow
计算机行业周报-全球AI产业跟踪:Nvidia发布6.3万亿Token大型AI训练数据库
Yong Xing Zheng Quan·2025-02-05 11:03

Investment Rating - The report maintains an "Accumulate" rating for the computer industry [6]. Core Insights - Nvidia has launched a large AI training database named Nemotron-CC, containing a total of 6.3 trillion tokens, with 1.9 trillion being synthetic data. The development utilized model classifiers and rephrasing techniques to ensure high quality and diversity of data [1][19]. - Adobe has introduced a series of new generative AI tools aimed at automating labor-intensive production tasks, including the "Firefly Bulk Create" application, which allows users to quickly adjust the size of up to 10,000 images or replace all backgrounds in one click [2][19]. - Salesforce's CEO stated that Microsoft Copilot has low usage rates and has not provided transformative experiences for customers, who prefer tools like ChatGPT [3][19]. - iFLYTEK has released the X1 deep reasoning model, which is the first domestically trained model with deep thinking and reasoning capabilities based on a fully domestic computing power platform. It claims to achieve industry-leading results with lower computing power [4][20]. Investment Recommendations - The report suggests that with Nvidia's training database and iFLYTEK's deep reasoning model, along with Adobe's new generative AI tools, the AI industry chain is expected to see continued improvement. It recommends focusing on investment opportunities in the IaaS layer, specifically mentioning companies like Microsoft (MSFT), Google (GOOGL), Meta (META), and Amazon (AMAZON) [5][21].