黄仁勋否认：不切实际

Core Viewpoint - NVIDIA's CEO Jensen Huang dismissed concerns about the company's advanced AI chips being transferred to China, stating there is no evidence of such activities and emphasizing the complexity and scale of NVIDIA's hardware [1]. Group 1: NVIDIA's AI Chips and Export Regulations - Huang highlighted that NVIDIA's systems are large and complex, making it impractical to secretly reroute them [1]. - The company’s clients are aware of export regulations and are committed to compliance, ensuring continued demand for NVIDIA's technology [1]. - The U.S. government recently lifted restrictions on high-end AI chip exports to China, marking a significant policy shift aimed at expanding U.S. technological influence [1]. Group 2: Changes in Product Strategy - Following the ban on the HGX H20 product, NVIDIA's next-generation AI accelerator will not be based on the Hopper architecture, with a shift towards GDDR7 design [3]. - The U.S. government has effectively prohibited the export of NVIDIA's H20 and AMD's Instinct MI308 AI chips to China, resulting in a loss of $5.5 billion for NVIDIA [3]. - NVIDIA's high-end accelerators H100 and H200 were banned from export to China before their launch, leading to the introduction of the H800, which was also banned shortly after [3]. Group 3: Future Developments and Market Competition - NVIDIA is reportedly planning to launch a GDDR7-based Hopper GPU, although the Hopper architecture was initially designed for HBM memory [4]. - The uncertainty surrounding NVIDIA's product offerings may increase the popularity of Huawei's Ascend accelerators in China, despite their inferior performance compared to NVIDIA's solutions [5]. - Observing how NVIDIA navigates regulatory complexities and develops future solutions for the Chinese market will be crucial, as the company aims to maintain its market share against competitors like Huawei [5].