不止GPU!英伟达(NVDA.US)GTC 2026连发LPU、CPU新品 全面布局AI数据中心每一环
NvidiaNvidia(US:NVDA) 智通财经网·2026-03-17 04:08

Core Viewpoint - Nvidia has officially launched multiple new chips and platforms at the GTC conference, including the Nvidia Groq 3 language processing unit (LPU) and the Vera central processing unit (CPU) cabinet, aimed at enhancing its competitive edge in the AI market [1] Group 1: Product Launches - Nvidia introduced five large server cabinets tailored for different scenarios within AI data centers [1] - The Nvidia Groq 3 chip, acquired through a $20 billion deal, focuses on AI inference, marking Nvidia's entry into dedicated inference chips to meet the urgent demand for AI model application [1] - The Groq 3 LPU platform integrates 128 independent Groq 3 LPUs, significantly enhancing throughput and revenue potential [2] Group 2: Strategic Developments - The LPX architecture is optimized for trillion-parameter models and million-token contexts, maximizing efficiency in power consumption, memory, and computing power [2] - The independent deployment of the Vera CPU cabinet, featuring 256 liquid-cooled Vera chips, signifies Nvidia's strategic shift towards a "GPU + CPU" dual-drive strategy, targeting the data center market dominated by Intel and AMD [3] Group 3: Market Impact - Nvidia's new product line is expected to significantly boost data center revenue, with projected revenue of $193.5 billion for fiscal year 2026, up from $116.2 billion in fiscal year 2025 [4] - Major tech companies, including Amazon, Google, Meta, and Microsoft, plan to invest $650 billion in AI capital expenditures this year, indicating a substantial market opportunity for Nvidia [4]