美股研究社·2025-10-20 11:46Core Viewpoint - Major tech companies are increasingly moving towards in-house chip manufacturing, which may threaten Nvidia's profit margins as they aim to capture a significant share of the AI chip market [3][4]. Group 1: Actions by Major Tech Companies - OpenAI has announced a partnership with Broadcom to design custom chips, indicating a shift towards in-house chip development [3]. - Meta plans to acquire chip startup Rivos to enhance its internal chip research and development [3]. - Amazon's Project Rainier is progressing well, with plans to deploy hundreds of thousands of its custom Trainium2 chips for AI workloads [3]. Group 2: Market Share Projections - By 2028, custom chips designed by companies like Google, Amazon, and OpenAI are expected to capture 45% of the AI chip market, up from 37% in 2024 and 40% in 2025 [4]. - Nvidia and its competitors, such as AMD, will retain the remaining market share [4]. Group 3: Competitive Dynamics - The "Magnificent Seven" tech giants are motivated to design their own chips to avoid dependency on Nvidia's monopoly, as Nvidia's chips are costly, impacting profit margins for cloud service providers [6]. - Google has begun selling its Tensor Processing Units (TPUs) to external cloud service providers, marking a direct competition with Nvidia [7]. Group 4: Development Stages of Chip Initiatives - Google has been developing TPUs for over a decade, making it a leader in the field, while Amazon and Microsoft are at different stages of their custom chip development [8]. Group 5: Long-term Implications for Nvidia - Analysts suggest that while Nvidia currently leads the market, the rise of custom chips will gradually erode its profit margins, akin to a "boiling frog" scenario [9]. - Nvidia's CEO downplays the competition from custom chips, emphasizing the company's comprehensive server systems rather than just individual GPUs [10]. Group 6: Market Growth and Demand - Despite the rise of custom chips, analysts believe that the overall AI chip market is large enough to accommodate growth for both Nvidia and its competitors [11]. - Nvidia has invested $47 billion in AI-related ventures from 2020 to September 2025, indicating its commitment to expanding its market presence [11]. Group 7: Challenges in Custom Chip Development - The complexity of developing custom chips means not all companies will succeed, which may mitigate the competitive threat to Nvidia [13].