谷歌对外销售芯片：博通大涨，英伟达AMD应声下跌

Core Viewpoint - Google is intensifying competition with Nvidia by selling its Tensor Processing Units (TPUs) to clients for use in their own data centers, marking a significant shift in its business strategy [2][3]. Group 1: Market Dynamics - Google is negotiating with companies like Meta Platforms to utilize its Tensor AI chips, which could threaten Nvidia's market dominance [2]. - Meta is considering purchasing Google TPUs worth billions starting in 2027 and renting TPU capacity from Google Cloud as early as 2026 [2]. - Following the news, Google's stock rose over 2% in after-hours trading, while Nvidia and AMD saw declines [3]. Group 2: Technological Advancements - Google's latest TPU v7 accelerator shows significant performance improvements, with each Ironwood TPU providing 4.6 petaFLOPS of dense FP8 performance, slightly surpassing Nvidia's B200 [5][6]. - The Ironwood architecture allows for the connection of up to 9216 individual chips, enabling a total bandwidth of 9.6 Tbps, which is crucial for large-scale computing [7][8]. - The system's reliability is highlighted by a reported uptime of approximately 99.999% since 2020, equating to less than six minutes of downtime annually [8]. Group 3: Competitive Landscape - Google’s TPU pods can scale significantly, with the latest generation capable of supporting up to 9216 chips, which is a substantial increase from previous models [15]. - The competition is intensifying as companies like Anthropic plan to utilize up to one million TPUs for their next-generation models, indicating a shift in the AI model training landscape [15][16]. - Analysts are increasingly questioning the impact of AI-specific ASICs on Nvidia's GPU dominance, as companies like Google and Amazon enhance their hardware capabilities [16].