Core Viewpoint - Nvidia is preparing to launch a new AI chip based on Groq technology specifically for the Chinese market, marking a strategic shift in its product offerings to address geopolitical challenges [1][3][4] Group 1: Product Strategy - The new chip for China will not be a downgraded version, differing from previous approaches where products were specifically designed with reduced capabilities for the Chinese market [1][4] - This new chip is designed to be compatible with other systems and is expected to be officially launched in May this year [4] - Nvidia plans to utilize Groq's chip for the "inference" phase of AI, which involves real-time responses such as answering questions and executing tasks [1][4] Group 2: Competitive Landscape - Despite Nvidia's dominance in the AI training market, it faces increasing competition in the inference market, with several Chinese AI giants developing their own inference chips [3][6] - Groq's chip features an SRAM memory architecture with a memory bandwidth of up to 150TB/s, which is seven times that of the Rubin GPU, optimized for low-latency token generation [6]

知情人士向路透社透露，英伟达正在为中国市场准备的芯片并非性能降级版，也不是专门为中国市场设 计的。这款新型芯片能够适配其他系统，预计将于5月上市。 英伟达计划将Groq的芯片用于所谓的"推理"环节，即让AI系统为用户回答问题、编写代码或执行任务。 在本周展示的产品中，英伟达计划将其即将推出的Vera Rubin芯片与Groq芯片结合使用。尽管英伟达在 AI训练市场占据主导地位，但在推理市场却面临着更为激烈的竞争。多家中国主要企业已开始生产自 己的推理芯片，其中包括百度这样的AI头部公司。 凤凰网科技讯 北京时间3月18日，据路透社报道，两位知情人士周二称，英伟达公司正准备推出一款能 够在中国市场销售的Groq AI芯片版本。 英伟达在去年年底以170亿美元收购了AI芯片创业公司Groq，并在本周加州圣何塞举行的年度GTC开发 者大会上，展示了基于Groq芯片的全新产品线。 截至发稿，英伟达尚未就此置评。(作者/箫雨) ...