英伟达GPU，被嫌弃了

Core Viewpoint - OpenAI is dissatisfied with some of NVIDIA's latest AI chips and has been seeking alternatives since last year, indicating a potential shift in the relationship between these two prominent companies in the AI sector [2][3]. Group 1: OpenAI's Concerns - OpenAI's dissatisfaction stems from the performance of NVIDIA's hardware in providing timely responses for specific queries, particularly in software development and AI communication, which has led to a need for new hardware to meet approximately 10% of its future inference computing demands [3][8]. - OpenAI has explored partnerships with startups like Cerebras and Groq to obtain faster inference chips, but negotiations with Groq fell through due to NVIDIA's $20 billion licensing agreement with Groq [4][5]. Group 2: NVIDIA's Position - NVIDIA's CEO Jensen Huang has denied reports of a strained relationship with OpenAI, asserting that the company plans to invest up to $100 billion in OpenAI and that customers continue to choose NVIDIA for inference due to its performance and cost-effectiveness [3][5]. - NVIDIA has engaged with companies like Cerebras and Groq to explore potential acquisitions of SRAM chip technology, which is crucial for enhancing inference capabilities [10]. Group 3: Market Dynamics - The AI industry is witnessing a shift towards inference-focused chips, with OpenAI's efforts reflecting a broader trend where companies are prioritizing speed and efficiency in processing user requests [7][8]. - Competitors like Anthropic and Google benefit from using proprietary chips designed specifically for inference, which may provide them with performance advantages over NVIDIA's general-purpose AI chips [8].